nhs r&d hta programme - university of warwickwrap.warwick.ac.uk/49877/1/wrap_waugh_csii...
Post on 21-Jul-2020
5 Views
Preview:
TRANSCRIPT
Health Technology Assessm
ent 2004;Vol. 8: No. 43
Clinical and cost-effectiveness of continuous subcutaneous insulin infusion for diabetes
Clinical and cost-effectiveness of continuous subcutaneous insulin infusion for diabetes
JL Colquitt, C Green, MK Sidhu, D Hartwell and N Waugh
Health Technology Assessment 2004; Vol. 8: No. 43
HTAHealth Technology AssessmentNHS R&D HTA Programme
The National Coordinating Centre for Health Technology Assessment,Mailpoint 728, Boldrewood,University of Southampton,Southampton, SO16 7PX, UK.Fax: +44 (0) 23 8059 5639 Email: hta@soton.ac.ukhttp://www.ncchta.org ISSN 1366-5278
FeedbackThe HTA Programme and the authors would like to know
your views about this report.
The Correspondence Page on the HTA website(http://www.ncchta.org) is a convenient way to publish
your comments. If you prefer, you can send your comments to the address below, telling us whether you would like
us to transfer them to the website.
We look forward to hearing from you.
October 2004
How to obtain copies of this and other HTA Programme reports.An electronic version of this publication, in Adobe Acrobat format, is available for downloading free ofcharge for personal use from the HTA website (http://www.ncchta.org). A fully searchable CD-ROM isalso available (see below).
Printed copies of HTA monographs cost £20 each (post and packing free in the UK) to both public andprivate sector purchasers from our Despatch Agents, York Publishing Services.
Non-UK purchasers will have to pay a small fee for post and packing. For European countries the cost is£2 per monograph and for the rest of the world £3 per monograph.
You can order HTA monographs from our Despatch Agents, York Publishing Services by:
– fax (with credit card or official purchase order) – post (with credit card or official purchase order or cheque)– phone during office hours (credit card only).
Additionally the HTA website allows you either to pay securely by credit card or to print out yourorder and then post or fax it.
Contact details are as follows:York Publishing Services Email: ncchta@yps-publishing.co.ukPO Box 642 Tel: 0870 1616662YORK YO31 7WX Fax: 0870 1616663UK Fax from outside the UK: +44 1904 430868
NHS libraries can subscribe free of charge. Public libraries can subscribe at a very reduced cost of £100 for each volume (normally comprising 30–40 titles). The commercial subscription rate is £300 per volume. Please contact York Publishing Services at the address above. Subscriptions can only bepurchased for the current or forthcoming volume.
Payment methods
Paying by chequeIf you pay by cheque, the cheque must be in pounds sterling, made payable to York PublishingDistribution and drawn on a bank with a UK address.
Paying by credit cardThe following cards are accepted by phone, fax, post or via the website ordering pages: Delta, Eurocard,Mastercard, Solo, Switch and Visa. We advise against sending credit card details in a plain email.
Paying by official purchase orderYou can post or fax these, but they must be from public bodies (i.e. NHS or universities) within the UK.We cannot at present accept purchase orders from commercial companies or from outside the UK.
How do I get a copy of HTA on CD?
Please use the form on the HTA website (www.ncchta.org/htacd.htm). Or contact York PublishingServices (see contact details above) by email, post, fax or phone. HTA on CD is currently free of chargeworldwide.
The website also provides information about the HTA Programme and lists the membership of the variouscommittees.
HTA
Clinical and cost-effectiveness ofcontinuous subcutaneous insulininfusion for diabetes
JL Colquitt,* C Green, MK Sidhu, D Hartwell and N Waugh
Southampton Health Technology Assessments Centre, Southampton, UK
* Corresponding author
Declared competing interests of authors: none
Published October 2004
This report should be referenced as follows:
Colquitt JL, Green C, Sidhu MK, Hartwell D, Waugh N. Clinical and cost-effectiveness ofcontinuous subcutaneous insulin infusion for diabetes. Health Technol Assess 2004;8(43).
Health Technology Assessment is indexed in Index Medicus/MEDLINE and Excerpta Medica/EMBASE.
NHS R&D HTA Programme
The research findings from the NHS R&D Health Technology Assessment (HTA) Programme directlyinfluence key decision-making bodies such as the National Institute for Clinical Excellence (NICE)
and the National Screening Committee (NSC) who rely on HTA outputs to help raise standards of care.HTA findings also help to improve the quality of the service in the NHS indirectly in that they form a keycomponent of the ‘National Knowledge Service’ that is being developed to improve the evidence ofclinical practice throughout the NHS.
The HTA Programme was set up in 1993. Its role is to ensure that high-quality research information onthe costs, effectiveness and broader impact of health technologies is produced in the most efficient wayfor those who use, manage and provide care in the NHS. ‘Health technologies’ are broadly defined toinclude all interventions used to promote health, prevent and treat disease, and improve rehabilitationand long-term care, rather than settings of care.
The HTA programme commissions research only on topics where it has identified key gaps in theevidence needed by the NHS. Suggestions for topics are actively sought from people working in theNHS, the public, consumer groups and professional bodies such as Royal Colleges and NHS Trusts.
Research suggestions are carefully considered by panels of independent experts (including consumers)whose advice results in a ranked list of recommended research priorities. The HTA Programme thencommissions the research team best suited to undertake the work, in the manner most appropriate to findthe relevant answers. Some projects may take only months, others need several years to answer theresearch questions adequately. They may involve synthesising existing evidence or designing a trial toproduce new evidence where none currently exists.
Additionally, through its Technology Assessment Report (TAR) call-off contract, the HTA Programme isable to commission bespoke reports, principally for NICE, but also for other policy customers, such as aNational Clinical Director. TARs bring together evidence on key aspects of the use of specifictechnologies and usually have to be completed within a limited time period.
The research reported in this monograph was commissioned and funded by the HTA Programme onbehalf of NICE as project number 01/53/01. The authors have been wholly responsible for all datacollection, analysis and interpretation and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the referees for theirconstructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.
The views expressed in this publication are those of the authors and not necessarily those of the HTAProgramme, NICE or the Department of Health.
Editor-in-Chief: Professor Tom WalleySeries Editors: Dr Peter Davidson, Professor John Gabbay, Dr Chris Hyde,
Dr Ruairidh Milne, Dr Rob Riemsma and Dr Ken SteinManaging Editors: Sally Bailey and Caroline Ciupek
ISSN 1366-5278
© Queen’s Printer and Controller of HMSO 2004
This monograph may be freely reproduced for the purposes of private research and study and may be included in professional journals providedthat suitable acknowledgement is made and the reproduction is not associated with any form of advertising.
Applications for commercial reproduction should be addressed to NCCHTA, Mailpoint 728, Boldrewood, University of Southampton, Southampton, SO16 7PX, UK.
Published by Gray Publishing, Tunbridge Wells, Kent, on behalf of NCCHTA.Printed on acid-free paper in the UK by St Edmundsbury Press Ltd, Bury St Edmunds, Suffolk.
Criteria for inclusion in the HTA monograph seriesReports are published in the HTA monograph series if (1) they have resulted from work commissionedfor the HTA Programme, and (2) they are of a sufficiently high scientific quality as assessed by the refereesand editors.
Reviews in Health Technology Assessment are termed ‘systematic’ when the account of the search,appraisal and synthesis methods (to minimise biases and random errors) would, in theory, permit thereplication of the review by others.
T
Objectives: To assess the clinical and cost-effectivenessof continuous subcutaneous insulin infusion (CSII)compared with multiple daily injections (MDI) in thedelivery of intensive insulin therapy for the treatmentof diabetes mellitus.Data sources: Electronic databases, references ofretrieved articles and manufacturer submissions.Experts in the field were consulted.Review methods: For the systematic review of clinicaland cost-effectiveness, studies were assessed forinclusion according to predefined criteria by tworeviewers. Data extraction and quality assessmentwere undertaken by one reviewer and checked by asecond reviewer. Data on clinical effectiveness weresynthesised through a narrative review with fulltabulation of all eligible studies, with meta-analysisperformed where appropriate.Results: Twenty studies comparing CSII with MDIwere identified. Quality was generally poor. In adultswith Type 1 diabetes, glycated haemoglobin improvedby 0.61% (95% CI –1.29 to 0.07) in longer termstudies, although this improvement was smaller when astudy using bovine ultralente was excluded. A reductionin insulin dose with CSII of about 12 units per day(–11.90, 95% CI –18.16 to 5.63) was found in short-term studies, with smaller differences in longer termstudies. Body weight and cholesterol levels weresimilar between treatments. Hypoglycaemic events didnot differ significantly between CSII and MDI in mosttrials, but some found fewer events with CSII and onefound more hypoglycaemia and hypoglycaemic comawith CSII. There was no consistency between thestudies in patient preference, but progress has beenmade both with insulin pumps and injector pens sincethe publication of many of the older studies. Nodifference in glycated haemoglobin between CSII andMDI was found in pregnancy; one study found lessinsulin was required by patients with CSII, but twoother studies found no significant difference. One study
of adolescents found lower glycated haemoglobin andinsulin dose with CSII whereas a second study found nosignificant difference. In CSII analogue insulin wasassociated with lower glycated haemoglobin levels thansoluble insulin. No economic evaluations comparingCSII with MDI were identified. The estimatedadditional cost of CSII compared to MDI varies from£1091 per annum to £1680 per annum, according tothe make of the insulin pump and the estimated life ofthe device. These estimates include the costs for theinsulin pump, the consumables associated with deliveryof CSII, and an allowance for the initial educationrequired when patients switch from MDI to CSII. The largest component of the annual cost for CSII is the cost of consumable items (e.g. infusion sets).Conclusions: When compared with optimised MDI,CSII results in a modest but worthwhile improvementin glycated haemoglobin in adults with Type 1 diabetes.It has not been possible to establish the longer termbenefits of such a difference in glycated haemoglobin,although there is an expectation that it would bereflected in a reduction in long-term complications.More immediate primary benefits from CSII may beassociated with an impact on the incidence ofhypoglycaemic events and the dawn phenomenon, andgreater flexibility of lifestyle. However, there is limitedevidence on this, and information presented to offercontext on quality-of-life is based on testimonies fromthose patients who have had a positive experience ofCSII. The estimated cost to the NHS per year for CSII would be around £3.5 million in England andWales if 1% of people with Type 1 diabetes used CSII,£10.5 million for 3%, and £17.5 million for 5%.Further research should focus on wider benefits ofCSII, such as flexibility of lifestyle and quality of life, and on the psychological impact of wearing a device for24 hours every day. Research into the use of CSII inchildren of different ages is also needed.
Health Technology Assessment 2004; Vol. 8: No. 43
iii
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Abstract
Clinical and cost-effectiveness of continuous subcutaneous insulininfusion for diabetes
JL Colquitt,* C Green, MK Sidhu, D Hartwell and N Waugh
Southampton Health Technology Assessments Centre, Southampton, UK* Corresponding author
Health Technology Assessment 2004; Vol. 8: No. 43
v
Glossary and list of abbreviations ............. vii
Executive summary .................................... ix
1 Aim of the review ...................................... 1Subsidiary questions ................................... 1
2 Background ................................................ 3Introduction ............................................... 3Control of high blood glucose levels ......... 3Hypoglycaemia ........................................... 5Insulin treatment ........................................ 5Continuous subcutaneous insulin infusion ....................................................... 7Why are pumps little used in the UK? ....... 9Subsidiary questions ................................... 11
3 Clinical effectiveness .................................. 13Methods ...................................................... 13CSII versus MDI ......................................... 14Analogue versus soluble insulin in CSII ........................................................ 38Discontinuation rates ................................. 42
4 The patient’s perspective .......................... 45Caveats ....................................................... 45Reasons for switching to insulin by pump ..................................................... 46Hypoglycaemic events ................................ 46Control of blood glucose ............................ 47Flexibility of lifestyle .................................. 47Other comments ........................................ 48Disadvantages ............................................. 48Children and CSII ..................................... 49Common themes ........................................ 51Conclusions ................................................ 51
5 Economic analysis ...................................... 53Introduction ............................................... 53Literature review: economic evaluations and quality of life comparisons .................. 54Costs associated with CSII (versus MDI) ............................................... 55The benefits of CSII ................................... 59Economic models for Type 1 diabetes ....... 66Cost-effectiveness of CSII versus MDI ............................................................ 71
6 Implementation ......................................... 77Barriers to implementation ....................... 79
7 Discussion: analysis of uncertainties ......... 81Efficacy ....................................................... 81Comparators .............................................. 82Cost-effectiveness and cost per QALY .......................................................... 82Costs ........................................................... 83Duration of studies ..................................... 83Strengths and limitations of review ........... 83Research needs ........................................... 84
8 Conclusions ................................................ 85
Acknowledgements .................................... 87
References .................................................. 89
Appendix 1 Rapid review methods from the research protocol ......................... 97
Appendix 2 Sources of information, including databases searched and search terms ............................................... 101
Appendix 3 Quality assessment scale ............................................................ 103
Appendix 4 List of excluded studies ......................................................... 105
Appendix 5 List of recent abstracts ........... 107
Appendix 6 Summary of methodology: adults with Type 1 diabetes ...................................................... 109
Appendix 7 Summary of methodology: pregnancy ........................... 125
Appendix 8 Summary of methodology: adolescents .......................... 131
Appendix 9 Summary of methodology: analogue versus soluble insulin ............................................ 135
Appendix 10 Summary of results: adults with Type 1 diabetes ........................ 141
Appendix 11 Summary of results: pregnancy ................................................... 153
Contents
Appendix 12 Summary of results: adolescents ................................................. 161
Appendix 13 Summary of results: analogue versus soluble insulin .................. 165
Appendix 14 Calculations for NHS staff costs ............................................................ 171
Health Technology Assessment reportspublished to date ....................................... 173
Health Technology Assessment Programme ................................................ 183
Contents
vi
GlossaryTypes of diabetes The classifications ofdiabetes have changed in recent years, with theold terminology of insulin-dependent and non-insulin dependent being replaced by Type 1and Type 2. This was because many patientswith non-insulin-dependent diabetes are nowtreated with insulin. The new classification isbased on the aetiology and pathology ofdiabetes, rather than on treatment.
Type 1 diabetes Type 1 diabetes indicates thatthere has been a process of destruction of thebeta cells of the pancreas, leading to insulindeficiency. Eventually, insulin is required for
survival in order to prevent the development ofketoacidosis and death. There is usually aprocess of autoimmunity with autoantibodies,but these are not seen in all patients. The causeis unknown.
Type 2 diabetes Type 2 diabetes is morecommon, and is characterised by insulinresistance and insulin deficiency. Thedeficiency may be relative to insulin needs,rather than absolute, and there may be higherthan normal production of insulin at somestages. Type 2 diabetes is linked to overweightand obesity and to physical inactivity.
Health Technology Assessment 2004; Vol. 8: No. 43
vii
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Glossary and list of abbreviations
Technical terms and abbreviations are used throughout this report. The meaning is usually clear fromthe context, but a glossary is provided for the non-specialist reader. In some cases, usage differs in the
literature, but the term has a constant meaning throughout this review.
List of abbreviationsBDR background diabetic retinopathy
BMI body mass index
CI confidence interval
CRD Centre for Reviews andDissemination
CSII continuous subcutaneous insulininfusion, delivered with the aid ofan insulin pump
CT conventional therapy. In Type 1diabetes, conventional therapy inthe UK usually means that patientstake twice daily injections ofmixtures of short-acting andintermediate or long-acting insulin.
DC day case
DCCT Diabetes Control and ComplicationsTrial
DKA diabetic ketoacidosis, a life-threatening metabolic disturbancerelated to shortage of insulin. It isoften brought on by other illnessessuch as infections, which increasethe body’s insulin needs
DQOL diabetes quality of life
DSN diabetes specialist nurse
DTSQ Diabetes Treatment SatisfactionQuestionnaire
ELIP elective inpatient cost
FBR fixed basal rate
FCE finished clinical episode, i.e. ahospital admission to one unit
HbA1c glycated haemoglobin A1c
HFS Hypoglycaemia Fear Survey
continued
Glossary and list of abbreviations
viii
List of abbreviations continued
HOR higher overnight rate
HRG health-related resource group
hypo hypoglycaemic event
IDDM insulin-dependent diabetes mellitus
INPUT voluntary organisation promotingInsulin Pump Therapy
ITT intention-to-treat
MDI multiple daily injections
NELIP non-elective inpatient cost
NICE National Institute for ClinicalExcellence
p = ns not statistically significant
OR odds ratio
PDR proliferative diabetic retinopathy
PUMP Pump Management for Professionals
QALY quality-adjusted life-year
QoL quality of life
QWB quality of well-being scale
RCT randomised controlled trial
SCL-90R Symptom Checklist-90R
SD standard deviation
SF-36 Short Form with 36 Items
SG standard gamble
TTO time trade-off
UKPDS United Kingdom ProspectiveDiabetes Study
VAS visual analogue scale
WESDR Wisconsin Epidemiologic Study ofDiabetic Retinopathy
WMD weighted mean difference
All abbreviations that have been used in this report are listed here unless the abbreviation is well known (e.g. NHS), or it has been used only once, or it is a non-standard abbreviation used only in figures/tables/appendices in which case the abbreviation is defined in the figure legend or at the end of the table.
Health Technology Assessment 2004; Vol. 8: No. 43
ix
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Description of the proposedserviceThis systematic review examines the clinical andcost-effectiveness of continuous subcutaneousinsulin infusion (CSII) using insulin pumpscompared with multiple daily injections (MDI) fordiabetes.
Epidemiology and backgroundThere are two main types of diabetes. Type 1diabetes involves a process of destruction of thebeta cells of the pancreas, leading to severe insulindeficiency, so that insulin treatment is required forsurvival. It represents about 10–15% of all diabetesin England and Wales. Type 2 diabetes is muchmore common, and is characterised by insulinresistance and relative insulin deficiency. Type 2diabetes is linked to overweight and obesity and tophysical inactivity. The number of people withinsulin-treated diabetes has increased owing to themarked increase in the incidence of Type 1diabetes and also to a greater number of peoplewith Type 2 diabetes being treated with insulin toimprove diabetic control. There has also been anincrease in the prevalence of Type 2 diabetes,particularly among the Asian community. Poorcontrol of diabetes, reflected in high blood glucose levels, can in the short term result indiabetic ketoacidosis, a serious and potentiallyfatal condition, and in the long term increase the risk of complications such as diabeticretinopathy and nephropathy. However, studieshave shown that good diabetic control isassociated with a reduced risk of thesecomplications.
If insulin levels are too high and blood glucosefalls, hypoglycaemic episodes occur. The effects ofa hypoglycaemic episode depend on how low theblood glucose level falls, varying from mild andrapidly corrected by food or sugary drinks, tosevere where help is required. Severehypoglycaemia can lead to unconsciousness,convulsions or death.
There are several problems with currenttreatment. In the non-diabetic state, the body
needs a little insulin all the time (basal insulin)boosted by increased output after meals. This isdifficult to achieve with conventional insulininjections, and in particular good control of bloodglucose during the night is difficult. Intensiveinsulin regimens such as CSII aim to resemblemore closely the output of a normal pancreas byproviding basal insulin for fasting periods andadditional short-acting supplements to covermeals.
MethodsA systematic review of the literature and aneconomic evaluation were undertaken.
Data sourcesElectronic databases were searched, including theCochrane Library, MEDLINE, EMBASE, PubMed,Science Citation Index, Web of ScienceProceedings, DARE and HTA databases,PsycINFO, CIHAHL, NHS Economic EvaluationDatabase, EconLIT and Health ManagementInformation Consortium database. References ofall retrieved articles were checked for relevantstudies and experts were contacted for advice andpeer review and to identify additional publishedand unpublished references. Manufacturersubmissions to the National Institute for ClinicalExcellence (NICE) were reviewed.
Study selectionStudies were included if they fulfilled the followingcriteria:
� Interventions: CSII using insulin pumpscompared with optimised MDI (at least threeinjections per day). Analogue compared withsoluble insulin in CSII.
� Participants: people with insulin-treateddiabetes (Type 1 or Type 2). Newly diagnosedpatients were excluded.
� Outcomes: glycated haemoglobin, insulin dose,weight change, lipid levels, patient preference,quality of life, adverse effects.
� Design: parallel randomised controlled trials(RCTs) and randomised and non-randomisedcrossover studies with a minimum duration of10 weeks on each treatment.
Executive summary
x
Studies in non-English language or available only as abstracts were excluded from the main analysis.
For questions where no eligible studies wereidentified, information from selected observationalstudies was discussed.
Titles and summaries of studies being assessed forinclusion were checked by two reviewers. Full textsof selected studies were assessed for inclusion byone reviewer and checked by a second. Differencesin opinion were resolved through discussion.
Data extraction and quality assessmentData extraction and quality assessment wereundertaken by one reviewer and checked by asecond reviewer, with any disagreement resolvedthrough discussion. The quality of includedstudies was assessed in accordance with the Centrefor Reviews and Dissemination (CRD) Report 4quality assessment scale.
Data synthesisData on the clinical effectiveness of CSII fordiabetes were synthesised through a narrativereview with full tabulation of results of all eligiblestudies, with meta-analysis performed whereappropriate. Cost-effectiveness analysis examinedthe marginal costs of CSII compared with MDIand considered evidence on the marginal benefitssuch as improved control, adverse events andquality of life.
Number and quality of studiesSearching identified 20 studies comparing CSIIwith MDI. These included eight parallel RCTs,nine randomised crossover studies and three non-random crossover studies. Fourteen studiesincluded adults with Type 1 diabetes, four studiesincluded pregnant women and two studiesincluded adolescents. The quality of reporting andmethodology of the studies, many of which datedfrom many years ago, were often poor by today’sstandards, with just two studies having adequaterandomisation and none reporting adequateallocation concealment.
No RCTs or crossover studies were identified inchildren, on overnight use of CSII, in patientswith poorly controlled Type 2 diabetes or ondiscontinuation rates; therefore, selectedobservational studies were discussed in thesesections.
Six studies (one parallel RCT and five random crossover studies) were identifiedcomparing analogue with soluble insulin in CSII. Randomisation and allocationconcealment were adequate in the parallel RCT but not reported in the crossover studies.
No economic evaluations comparing CSII withoptimised MDI were found.
Summary of benefitsAdults with Type 1 diabetesIf all trials were included, a mean improvement in glycated haemoglobin of about 0.6% was found with CSII compared with MDI in both short-term [–0.64, 95% confidence interval (CI)–1.28 to 0.01] and longer term (–0.61, 95% CI–1.29 to 0.07) studies. This improvement wassmaller if a study which used bovine ultralente inthe control arm was excluded; the reduction inglycated haemoglobin A1c (HbA1c) is then only0.5%. Short-term studies show a reduction ininsulin dose of about 12 units (–11.90, 95% CI–18.16 to –5.63), with less difference in longer term studies. Body weight was similarduring treatment with CSII and MDI. The twostudies that reported data on cholesterol levelsfound no significant difference between thetreatments. There was no consistency between thestudies in patients preferring CSII or MDI,although many of the older studies used older,bulkier and less reliable pumps, and progress hasalso been made with discreet ‘pen’ injectors inMDI; therefore, these findings are probably notrelevant to the present devices. Hypoglycaemicepisodes did not differ significantly between CSII and MDI in most trials, but some foundfewer episodes with CSII and one study found more hypoglycaemia and hypoglycaemiccoma with CSII. In some observational studies, much greater reductions in the number of severe hypoglycaemic episodes were seen with CSII, which may be because these studiestend to select patients having particular problems.
PregnancyThree studies found no difference in glycatedhaemoglobin between CSII and MDI. Less insulin per kilogram was required by patients with CSII in one study, but two otherstudies found no significant difference. Patient preference and quality of life were notreported.
Executive summary
AdolescentsOne study found no significant difference between CSII and MDI, whereas the second studyfound lower glycated haemoglobin and insulindose with CSII. Over half of the patients chose tocontinue treatment with CSII in the former study.
ChildrenNo randomised trials were identified. Case seriessuggest that CSII has a place in treatment ofchildren with diabetes, but this needs to beconfirmed in randomised studies.
Overnight only CSIIThe combination of overnight CSII and daytimeMDI may help in children, by reducing nocturnalhypoglycaemic episodes and the dawnphenomenon, but no randomised trials wereidentified, and further research is necessary.
Short-term use in adults with poorlycontrolled Type 2 diabetesIt has been suggested that short-term CSII mayhelp in patients with Type 2 diabetes on highdoses of oral drugs and who are resistant toinsulin. No good evidence was found.
Analogue versus soluble insulinIn CSII, analogue insulin was associated withlower glycated haemoglobin levels than solubleinsulin and was preferred by patients. Nodifference in insulin dose or weight change wasobserved. Some studies found fewer hypoglycaemic episodes with analogue insulin,although this varied according to the definitionsused.
CostsThe additional cost of CSII compared with MDIvaries according to the make of pump and theestimated life of the device, from £1091 perannum using the cheapest pump and assuming an8-year life of the pump to £1680 per annum withthe most expensive model and assuming a life ofonly 4 years. These estimates include costs forconsumables and the initial education requiredwhen patients switch from MDI to CSII. Thelargest component of cost is the consumable items,such as infusion sets (tubing, etc.), with the capitalcost of the pump secondary. Initial education forthose switching to CSII is very important, and weestimated an additional cost per patient switching from MDI to CSII to be in the region of £150.
Costs per life year gainedThere are definite benefits of CSII over MDI,including improved control of diabetes, not just asreflected in glycated haemoglobin and in a slightlyreduced incidence of severe hypoglycaemic events,but also in flexibility of lifestyle and hence qualityof life. However, evidence on quality of life isreported in only one trial, and comes mainly fromtestimonies of pump users.
One would expect the improvement in HbA1c tobe reflected in reduced long-term complicationsand for that to be accompanied by reduced coststo the NHS. However, we have not found asatisfactory method of converting the observedbenefits into a cost per quality-adjusted life-year.
The main problem with the current evidence isthat it does not fully reflect the selection ofpatients for CSII. Most people on insulin therapywould not have much to gain from CSII, but thosewith particular problems such as recurrent severehypoglycaemia would. Their benefits wouldinclude not only fewer hypoglycaemic episodes,but also a reduction in fear of them. However, theutility effect of the reduction in fear ofhypoglycaemic episodes has not been quantified.The cost-effectiveness of CSII is likely to be muchbetter for certain subgroups.
Sensitivity analysisThe main costs are of consumables and pumps.The price of pumps might come down with bulkpurchase, but this is speculative. This would nothave much impact on the cost per annum.
ConclusionsControl of diabetes consists of more than justcontrol of blood glucose as reflected in glycatedhaemoglobin. Compared with optimised multipleinjection insulin therapy, CSII results in a modestbut worthwhile improvement in glycatedhaemoglobin, but its main value may be in reducing other problems such as hypoglycaemiaand the dawn phenomenon, and in improvingquality of life by allowing greater flexibility oflifestyle. Pumps appear to be a useful advance forpatients having particular problems, rather than adramatic breakthrough in therapy, and wouldprobably be used by only a small percentage ofpatients.
Health Technology Assessment 2004; Vol. 8: No. 43
xi
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Implications of approval of anincreased use of CSIIMany health authorities are not funding insulinpumps, and some of those that are have restrictedthe number. Many patients are funding their ownpumps. According to clinical consensus, it isunlikely that CSII would be used by more than asmall proportion of people with Type 1 diabetes,but the exact proportion is not known. We wouldnot expect any use in true Type 2 diabetes in theforeseeable future. The cost to the NHS per yearwould be around £3.5 million in England and Walesif 1% of people with Type 1 diabetes used CSII,£10.5 million for 3% and £17.5 million for 5%. Theeducational needs of patients starting CSII aresignificant, and it would usually be diabetesspecialist nurses who would provide this. However,there are many other demands on their time.
Need for further researchThe trials to date have focused on easilymeasurable outcomes such as glycatedhaemoglobin. The main benefits may be in termsof flexibility of lifestyle and quality of life, anddata on those would help with cost-effectivenessanalysis. Some of the implications for patientssuch as the psychological impact of wearing adevice for 24 hours every day have not beenquantified.
There appears to be no wholly satisfactoryeconomic model for diabetes, which would allowimprovements in diabetes control to be convertedinto a cost per quality-adjusted life-year. Researchis also needed into the use of CSII in children ofdifferent ages.
Executive summary
xii
The main aim of this review is to assess theclinical effectiveness and cost-effectiveness of
continuous subcutaneous insulin infusion (CSII)using insulin pumps, compared to intensivetreatment with multiple daily injections (MDI).The benefits could be improved control of bloodglucose levels as reflected in glycated haemoglobinA1c (HbA1c), or a similar level of control but withother advantages such as fewer problems withhypoglycaemic episodes (hypos), or greaterflexibility of lifestyle and hence better quality oflife (QoL).
Both CSII and MDI are forms of intensive insulintreatment. The benefit of intensive treatment, forthose whose diabetes is inadequately controlled onconventional insulin therapy, is not in doubtfollowing trials such as The Diabetes Control andComplications Trial (DCCT) and the UnitedKingdom Prospective Diabetes Study (UKPDS).Conventional insulin therapy (CT) is assumed to be twice daily injections usually of a mixture of short-acting and long-acting insulins. Thisreport assumes that Type 1 patients wouldnormally try twice-daily mixtures, and would then be treated with MDI before trying CSII, thatis, patients would not go directly fromconventional to CSII. However, increasingly,patients may start on MDI from diagnosis. Type 2patients would usually start with a once-dailyinjection of a long-acting insulin (ultralente or along-acting analogue such as glargine) or twice-daily NPH, and if not well-controlled wouldprogress next to CT, before going on, if necessary,to MDI and then CSII. This report thereforeexcludes the use of CSII in newly diagnosedpatients.
Intensive treatment with MDI is taken to meantreatment with a combination of short-actinginsulins (soluble or analogue) to cover meals, with long-acting insulin (intermediate, ultralenteor very long-acting analogues such as glargine) to provide basal insulin. In practice, MDI willmean a minimum of three injections a day, butmay involve more than that. It will also involveblood glucose testing. There will be varying levels
of intensity within MDI, and so the comparator forCSII is optimised MDI.
It should be noted that treatment of Type 1diabetes involves more than just injection ofinsulin:
“Modern insulin therapy is a package of insulinadministration according to basal and bolusprinciples, frequent blood glucose self monitoring,insulin adjustment according to blood glucoseresults and meal size, exercise, plus intake of ahealthy diet, and full diabetes education.” (Pickup J, Guy’s, King’s and St Thomas’ School ofMedicine, London: personal communication,2002)
Subsidiary questionsA number of subsidiary questions are alsoaddressed. Where eligible RCTs and crossoverstudies are lacking, evidence from selectedobservational studies will be sought.
1. Does CSII have advantages over MDI inwomen with Type 1 diabetes pre-conceptuallyand during pregnancy?
2. What is the role of CSII in adolescents andchildren?
3. What is the role of overnight-only CSII(injections taken as usual during the day)?
4. What is the value of short periods of CSII inpatients with Type 2 diabetes who are veryresistant to oral drugs such as thesulphonylureas?
5. In CSII, do analogue insulins have advantagesover older short-acting (soluble) insulins?
6. How many patients continue to use CSII(discontinuation rates)?
Implantable pumps are not covered by the review.
Insulin pumps are used for insulin therapy insome patients in hospital, for example thosehaving surgery. Such very short-term use is notrelevant to this review.
Health Technology Assessment 2004; Vol. 8: No. 43
1
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Chapter 1
Aim of the review
IntroductionThe number of people with insulin-treateddiabetes has increased over recent decades for tworeasons. Firstly, there has been a marked increasein the incidence of Type 1 diabetes mellitus[formerly called insulin-dependent diabetesmellitus (IDDM)].1–7 In some parts of the UK, theincidence of Type 1 diabetes has trebled over thelast 30 years.1,5 Rates in the UK are high ininternational terms,8 although more so inScotland. Second, more people with Type 2diabetes (formerly non-insulin-dependent diabetesmellitus) are being treated with insulin in order toachieve better control of blood glucose levels. Thistrend started before, but was probably acceleratedby the results of, the UKPDS.9 There has also beenan increase in the prevalence of Type 2diabetes,10,11 particularly among those from ethniccommunities.12 A recent study by the Office forNational Statistics shows a rise in prevalence from1994 to 1998.13
The need for insulin arises for different reasons inthe two main types of diabetes. Insulin is producedby the islet cells of the pancreas. In Type 1diabetes, there is a true insulin deficiency becausethe islet cells are destroyed as the result of anautoimmune process, for cause or causes unknown.In Type 2 diabetes, it is more common to have anormal or increased insulin production initially,with insulin resistance being a feature. However,over the years the pancreas may fail, leading toinsulin deficiency. Insulin production rates willhave fallen in patients to about 50% by the time ofdiabetes diagnosis in the UK, as was shown in oneof the UKPDS papers.14 It is probable that thisprogressive worsening would be much less, ormight not occur, if patients adhered to lifestylemeasures such as weight loss and exercise.
There are some patients who do not fit neatly intothe two groups, because at diagnosis they appearto have Type 2 diabetes, but then progress toinsulin treatment over the course of a few years;they may have slow onset Type 1 diabetes.15,16
The approximate prevalences of insulin-treateddiabetes in different age groups are shown inTable 1.
Control of high blood glucose levelsIf insulin levels are too low, blood glucose levelsare too high and there are other accompanyingmetabolic abnormalities. In the long run, poorcontrol of blood glucose increases the risk ofcomplications such as eye disease (diabeticretinopathy), which can potentially lead to loss ofvision. Diabetic retinopathy is the single mostcommon cause of blindness in people in theworking years of life in the UK and otherindustrial countries, and kidney disease (diabeticnephropathy) is now the leading cause of renalfailure, leading to the need for dialysis andtransplantation. Amputation rates are high indiabetes owing to the combination of peripheralvascular disease and neuropathy, and the care ofthose with severe peripheral vascular diseaseleading to amputation is one of the mostexpensive complications of diabetes.15 In the shortterm, poor diabetic control in Type 1 diabetes canresult in the metabolic derangement known asdiabetic ketoacidosis, which is a serious andoccasionally fatal condition, often referred to as‘diabetic coma’, because it can lead to reducedconsciousness and coma. Heart disease is morecommon amongst people with diabetes, but is lessclearly linked with blood glucose control.
Both randomised trials and large cohort studies haveshown that good control of blood glucose levels bytreatment is associated with reduced risk ofcomplications such as retinopathy and renal disease(older studies are in the meta-analysis by Wang andcolleagues,18 DCCT19 and Reichard andcolleagues20). For example, in the DCCT study, therewas a marked drop in the prevalence of serious eyedisease at 9 years, from 32% in the conventionaltreatment group to 9% in the intensive group.21 Inthis study, the intensive group maintained an averageHbA1c level of 7.2%, about 2% below those treatedconventionally (but note that the conventional groupwas better controlled than the US average). Therehas therefore been great emphasis on better(‘tighter’) control of blood glucose.
A meta-analysis by Egger and colleagues22
examined the risk of ketoacidosis during intensifiedtreatment compared with conventional treatment.For trials where patients could choose between
Health Technology Assessment 2004; Vol. 8: No. 43
3
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Chapter 2
Background
Background
4
TABLE 1 Prevalence of insulin treated diabetes per 1000 patients, by age, sex and year: 1994–98 in England and Wales
0–4 5–15 16–24 25–34 35–44 45–54 55–64 65–74 75–84 85+ Crude Age-rate standard-
(all ages) ised rate (all ages)
Males 1994Rate/1000 0.2 1.1 3.1 4.7 5.4 5.6 7.8 9.2 8.3 4.4 4.8 4.6LCL 0.1 0.9 2.7 4.3 4.9 5 7 8.3 7.1 2.7 4.6 4.4UCL 0.5 1.4 3.6 5.2 5.9 6.1 8.6 10.1 9.6 6.7 4.9 4.8No. of cases 7 86 171 388 393 375 390 372 163 20 2365 2365
Males 1995Rate/1000 0.2 1.4 3 4.5 5.5 5.8 8.3 9.7 9.4 5.9 5 4.8LCL 0.1 1.2 2.6 4.1 5 5.3 7.5 8.7 8.1 4 4.8 4.6UCL 0.5 1.7 3.5 5 6 6.4 9 10.6 10.7 8.5 5.2 4.9No. of cases 7 112 169 380 421 420 437 406 200 29 2581 2581
Males 1996Rate/1000 0.2 1.5 3.2 4.6 5.7 6.2 8.9 10 9.9 5.8 5.2 5LCL 0.1 1.2 2.7 4.2 5.2 5.7 8.1 9.1 8.6 3.9 5 4.8UCL 0.4 1.8 3.7 5.1 6.2 6.8 9.7 11 11.2 8.2 5.4 5.2No. of cases 7 125 184 401 463 481 495 439 227 30 2852 2852
Males 1997Rate/1000 0.3 1.6 3.3 4.7 5.8 6.9 9.5 11.4 10.2 6.9 5.6 5.3LCL 0.1 1.3 2.8 4.3 5.3 6.3 8.7 10.4 8.9 4.9 5.4 5.1UCL 0.6 1.9 3.7 5.2 6.3 7.4 10.2 12.3 11.4 9.4 5.8 5.5No. of cases 10 139 196 418 495 554 556 515 246 38 3167 3167
Males 1998Rate/1000 0.2 1.7 3.5 4.6 6.2 7.2 10 13.3 10.9 6.8 6 5.7LCL 0 1.4 3 4.1 5.7 6.6 9.3 12.3 9.7 4.9 5.8 5.5UCL 0.4 1.9 3.9 5 6.7 7.7 10.8 14.3 12.2 9.2 6.2 5.8No. of cases 5 151 219 413 565 609 634 627 282 41 3546 3546
Females 1994Rate/1000 0.1 1.7 3.2 3.4 4.4 4.5 7.7 8.3 8 4.7 4.4 4.1LCL 0 1.4 2.7 3 3.9 3.9 7 7.5 7 3.6 4.2 3.9UCL 0.3 2 3.6 3.8 4.9 5 8.5 9.1 9 6.1 4.6 4.3No. of cases 3 117 160 270 309 289 383 396 253 59 2239 2239
Females 1995Rate/1000 0.4 1.6 3.5 3.7 4.6 4.6 7.9 9.2 8.4 4.9 4.7 4.3LCL 0.2 1.3 3 3.3 4.1 4.1 7.2 8.3 7.5 3.8 4.5 4.1UCL 0.6 1.9 4 4.1 5 5.1 8.7 10 9.4 6.2 4.9 4.5No. of cases 11 120 178 305 340 323 415 450 289 65 2496 2496
Females 1996Rate/1000 0.3 1.7 3.3 3.8 4.8 5.1 8 9.5 8.8 4.5 4.8 4.5LCL 0.1 1.4 2.8 3.4 4.3 4.6 7.2 8.6 7.8 3.5 4.7 4.3UCL 0.6 2 3.8 4.2 5.3 5.6 8.7 10.3 9.8 5.8 5 4.6No. of cases 9 133 174 323 376 381 436 476 320 64 2692 2692
Females 1997Rate/1000 0.4 1.8 3.4 3.9 5 5.3 8.8 10.8 9.5 5.2 5.2 4.8LCL 0.2 1.5 2.9 3.5 4.5 4.8 8 9.9 8.5 4.1 5 4.6UCL 0.6 2.1 3.9 4.3 5.5 5.8 9.6 11.7 10.5 6.6 5.4 4.9No. of cases 11 147 189 340 410 412 508 558 362 78 3015 3015
Females 1998Rate/1000 0.3 1.9 3.2 4.3 5.2 5.7 9.4 12.1 9.4 5.9 5.5 5.1LCL 0.1 1.6 2.8 3.8 4.7 5.2 8.6 11.1 8.4 4.7 5.3 4.9UCL 0.5 2.2 3.7 4.7 5.6 6.2 10.1 13 10.3 7.2 5.7 5.2No. of cases 8 165 193 380 452 473 582 644 372 93 3362 3362
LCL, lower confidence limit; UCL, upper confidence limit.Source: Office for National Statistics.17
multiple injections or pump treatment, there wassome evidence of an increased risk of ketoacidosiswith intensified treatment, although this did notreach statistical significance [odds ratio (OR) 1.28,95% confidence interval (CI) 0.90 to 1.83, p =0.17]. For trials that randomised patients to pumps,there was a substantial increase in the risk ofketoacidosis (OR 7.20, 95% CI 2.95 to 17.58, p <0.0001). However, in the latter meta-analysis themost recent study was conducted in 1992, thereforethese results may no longer apply to newer pumps.
HypoglycaemiaHowever, insulin treatment may lead to bloodglucose levels falling below normal, potentiallyrobbing the brain of its essential glucose supply.This is called hypoglycaemia. Hypoglycaemicepisodes are usually known as ‘hypos’. Theconsequences vary depending on how low theblood glucose level falls. Mild hypoglycaemicevents may cause only a feeling of hunger andsweating, and can be rapidly corrected by food orsugary drinks. However, they may reduce theamount and quality of sleep when they occurduring the night (nocturnal hypoglycaemicevents). More serious or sustained falls in glucoselevel may mean that the diabetic person needshelp in order to recover, and severe hypoglycaemicevents can lead to behavioural disturbances,unconsciousness, convulsions or death. In youngchildren with frequent or severe hypoglycaemicevents, there may be some impairment ofintellectual function.23
People with Type 1 diabetes, especially youngerones, may be more afraid of hypoglycaemic eventsthan of long-term complications. As one of ourexpert advisers commented;
“Even though any single hypo event is short-livedin terms of its acute physiological effect, thepsychological effect on many patients is not at allshort-lived. It often has a very profound effect sothat the patient will do everything they can to avoida recurrence. Many patients have a greater fear ofhypos than they have of developing diabetes-relatedcomplications, and as a result will keep their bloodglucose levels higher than recommended in orderto avoid hypos. If they lost their fear of hypos,better glycaemic control could be achieved resultingin a reduced risk for complications.” (Tieszen K,Hope Hospital, Salford: personal communication)
Egger and colleagues22 reported that the incidenceof severe hypoglycaemia ranged from 0 to 33 per
100 person-years amongst conventionally treatedpatients and from 0 to 66 in intensively treatedpatients, in a meta-analysis of 14 randomisedcontrolled trials (RCTs) comparing conventionalwith intensive insulin regimens. The risk of severehypoglycaemia was three times as high in theintensive arms of the trials. However, because ofthe relative sizes of the trials, the result wasdominated by the DCCT study,24 which provided459 of the 531 hypoglycaemic events in theintensive group and 255 of the 315 in theconventional group, in the meta-analysis. Takingout DCCT would leave only 72 and 60hypoglycaemic events in the intensive andconventional groups, respectively. The risk ofsevere hypoglycaemia in the intensively treatedgroup fell after the first 2 years of the DCCTfeasibility study, suggesting a learning experiencein the trialists, but there was no reduction overtime in the annual rate of hypoglycaemic events inpatients recruited to the main study, suggestingthat hypoglycaemia will always be a risk inintensive therapy. Over an average follow-up of6.5 years, 65% of patients in the intensive and 35%of those in the conventional group had at least onesevere hypoglycaemic event.
In a representative group of people with IDDM,Muhlhauser and colleagues25 found that 10% ofconventionally treated patients and 9% of CSIItreated patients had at least one severehypoglycaemic event per year. The intensivelytreated group successfully improved their controlof blood glucose as reflected in HbA1 without anincreased risk of severe hypoglycaemia. There isgood evidence that although all patients with Type 1 diabetes are at some risk, the risk of severehypoglycaemia tends to be concentrated in arelatively small subset of individuals. Rewers andcolleagues26 found that 79% of severehypoglycaemic events occurred in the 14% ofchildren who had recurrent hypoglycaemic events.Thus, the strongest risk indicator for severehypoglycaemia is a history of an episode in thepast and this led the DCCT to modify theirrecruitment protocol after the first year to excludesuch patients. Loss of the warning symptoms of ahypoglycaemic events (see below) and deliberateor uniformed misuse of insulin also contribute torisk.
Insulin treatmentInsulin is ineffective if given by mouth, because itwould be digested. Therefore, at present, it has tobe given by injection. Inhaled forms are being
Health Technology Assessment 2004; Vol. 8: No. 43
5
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
developed but bioactivity is low (10–20%) andvariable, and there have been problems withaccurate delivery.
There are many forms of insulin on the market,but they fall into a number of groupings byduration of action (Table 2).
There are several problems with current treatment.First, the body needs a little insulin all the time(‘basal insulin’, which accounts for about half of thetotal insulin production) with brief peaks ofincreased levels for an hour or two after meals.The basal insulin need is around 0.5–1.0 units/hourin adults. After meals blood levels normallyincrease 5–10-fold in the first 30 minutes,dropping back to basal after about 2 hours.27 It isdifficult to achieve this normal profile of insulinrelease with injected insulin. The rapid-actinganalogues are closest. Soluble insulin injectionsstart to act about 30 minutes after injection, rise toa peak of action over the next hour and fall 4–6hours later, although they can cause hypoglycaemiaup to 8 hours after injection, particularly in theearly part of the night. The newer analogueinsulins such as lispro and aspart start much morepromptly but are shorter lived in effect and still donot replicate the normal non-diabetic state. Iflarger doses of short-acting insulins are given tocontrol hyperglycaemia, which occurs 1–2 hoursafter meals, hypoglycaemic events may result at2–4 hours or even up to 8 hours later, particularlyduring the night.
Figure 1 illustrates some of the problems. In each,the dotted lines show the insulin requirements aftermeals, when in non-diabetic people there is a steepbut short-lived rise in insulin production. The darksolid lines show the timing of action of solubleinsulin and the light solid lines that of long-actinginsulin. The diagrams show that soluble insulin hasa longer but lower peak of action; it does not mimicnormal pancreatic function at all well, being
insufficient in the hour or so after the meal, butthen liable to have too much effect 3–4 hours afterthe meal. Giving two injections a day of a mixtureof short-acting and longer acting insulin fails tocontrol peaks of blood glucose (not shown, butrepresented by the dotted lines of insulin needs)but will provide too much insulin for much of theday (Figure 1a). Moving to MDI (several injectionsof short-acting insulin during the day with onelong-acting at night) still does not providesatisfactory cover (Figure 1b), nor does a yet morecomplicated regimen of three injections of short-acting and two of long-acting (Figure 1c). However,the MDI regimens do approximate more closely toinsulin needs. Using more rapid acting analogueinsulins (not shown, but shorter peaks than soluble)is an improvement but still does not replicatenormal pancreatic insulin production.
Background
6
TABLE 2 Insulin preparations
Duration of Type Time to onsetaction
Rapid-acting Analogues 15 minutes
Short-acting Soluble human 15–60 minutes
Intermediate NPH and human 2–4 hoursand porcine ultralentes
Long Long-acting analogues
Insulin replacement
Solubleinsulin
Intermediateinsulin
Insulin replacement – multipledaily injection therapy
Insulin replacement – multiple daily injection therapy with split basal insulin
(a)
(b)
(c)
FIGURE 1 Insulin profiles
Second, absorption rates from the injection sitesvary from day to day. Absorption is affected byinjection site and depth, adiposity, exercise,temperature, skin blood flow and insulinpreparation.28 Absorption varies in the sameindividual by about 25%.28 This can lead tounpredictable hypoglycaemic episodes, usuallyassociated with longer acting insulins.
Third, there is a problem with overnight control,when a low basal level is needed, but with a rise inthe morning to counteract the rise in bloodglucose which then occurs (before wakening).Isophane and lente insulins cannot maintainnormal blood glucose levels overnight withoutincreasing the risk of hypoglycaemic events.Nocturnal hypoglycaemic events may not besevere, but can be troublesome and disruptive ofsleep. If the dose is reduced to avoidhypoglycaemia, it is likely that glucose levels willbe high by morning. The rise in blood glucoselevels towards the end of the night and the loss ofcontrol which then occurs with long-acting insulininjections given the night before are sometimesreferred to as the ‘dawn phenomenon’.
Fourth, patients may not comply with advice, andit is known that some patients do not take all theirprescribed insulin.29 This may be partly because ofthe limitations placed on lifestyle by the diet andinsulin regimens still used in most places. Thelimitations are imposed by the need to have mealsat regular intervals. There may be other optionssuch as the intensive educational programme DoseAdjustment for Normal Eating (DAFNE), whichwill be covered by another National Institute forClinical Excellence (NICE) review.
The net effect of the limitations in conventionalinsulin therapy is that diabetes is not wellcontrolled in the majority of patients. In a recentaudit, only a minority of young people withdiabetes achieved good control.30
A recent review31 concluded that:
“A substantial advance in diabetes care would allowindividuals to have euglycaemia (normal bloodglucose) without risk of severe hypoglycaemia, andwith much less effort than is currently devoted tointensive therapy.”
Continuous subcutaneous insulininfusionThe first studies of CSII delivered via insulinpumps came from Guy’s Hospital, London, in
197832 and Yale in 1979.33 The aim of CSII is totry to approximate the insulin delivery profilemore closely to the pattern of output behaviour ofthe normal pancreas by providing continuouslyinfused, low-volume basal insulin for fastingperiods and the delivery of increased rate bolusesto cover meals. Only short-acting (crystalline orrapidly absorbed analogues) insulin is used.
There have been several reviews of CSII. Thorp,34
writing after a relatively short period of CSIIexperience in 1986, noted that: “Efforts toimprove insulin administration have resulted inthe development of a myriad of electromechanicalpump devices.” He noted that technical problemswere common. He also commented that 30% ofpatients in one of the studies discontinued use ofthe pump, and that: “All of those discontinuinguse of the pump identified the physician as thekey person involved in the decision to try thisform of treatment, rather than themselves.”
Davies and Baum in 198835 considered theevidence on CSII use in children, and noted a lackof enthusiasm amongst British paediatricians, butcommented that the recent advent of injectiondevices such as pens might be part of the answer,since those were being seen as an alternative wayof administering intensive insulin regimens.
Over the past decade, much effort has gone intoimprovement of both pump and insulintechnology and there have been several recentreviews. Lenhard and Reeves36 carried out areview, using MEDLINE only. They noted the risein popularity of CSII after the introduction ofpumps in the late 1970s and early 1980s, followedby a fall because of size, safety and efficacyconcerns, followed then by a rise in usage after thepublication of the DCCT study. They also notedthat the newer pumps, reduced to three in numberfrom the ‘myriad’ described by Thorp, weresmaller, more reliable and easier to use. Theyestimated that about 8% of all adults in NorthAmerica with Type 1 diabetes were using pumps.They concluded that there was good evidence forbenefits in adults (‘comparable or slightly superiorto MDI’), and some in pregnancy, but that therewas little good-quality evidence in children.
The annual position statement from the AmericanDiabetes Association for 200237 was cautious, andsaid that: “an insulin pump may provide greaterlifestyle flexibility, particularly with regard to mealschedules and travel but may be too demandingfor some individuals.” The statement did notmake any recommendation on the relative merits
Health Technology Assessment 2004; Vol. 8: No. 43
7
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
of CSII and MDI, but said that: “pump therapy isas safe as multiple-injection therapy whenrecommended procedures are followed.”
Pickup and Keen,38 who were the originators ofCSII, reviewed the history of and evidence basefor CSII.39 They note the considerable world-wideuse of pumps (over 200,000 patients) and thedisproportionately low UK use. They concludethat on CSII, blood glucose and HbA1c are similaror slightly lower than with MDI, thathypoglycaemia is much less frequent and thatketoacidosis occurs at the same rate. They proposea set of clinical guidelines for identifying whichpatients should be considered for CSII, andconclude that the proportion of patients whowould be suitable is relatively small. They observethat individual patient choice and preference isnot possible unless the CSII option is available fora comparative trial, a condition rarely met in theUK. In a complementary study, Pickup andcolleagues40 carried out a meta-analysis of RCTscomparing CSII with MDI. They found that HbA1c
was about 0.5% better on CSII, but found that fewstudies reported hypoglycaemic events; noneappeared to report effect on QoL. The CSII groupneeded 14% less insulin.
Three health technology assessment reports haveexamined CSII. The Catalan Agency for HealthTechnology Assessment and Research (CAHTA)41
concluded that CSII was no more effective thanMDI, but appears to have based this only on theDCCT study. However, in the DCCT study patientswere not randomised to pumps or MDI; it was leftto individual centres to decide, and it would beunsafe to draw any comparison of the relativemerits of CSII and MDI.
The report from the Spanish Health TechnologyAssessment Agency [Agencia de Evaluacion deTecnologies Sanitarias (AETS)] came to similarconclusions.42 The US Agency for Health CarePolicy and Research (AHCPR, 1990)43 report issuperseded by new evidence.
It is always interesting to know what treatmentsclinicians with diabetes choose for themselves. A survey44 by the American Association ofDiabetes Educators (AADE) and the AmericanDiabetes Association (ADA) asked members if theyhad diabetes and, if so, how they were treated.About 6.4% of members had diabetes, of whom72% had Type 1. The survey found that 96% ofthose with Type 1 used an intensive insulinregimen, and that over half (60% of the AADEmembers with diabetes and 52% of the ADA
ones) used an insulin pump and, anecdotally, anumber of diabetic specialist physicians are known to have been on CSII for about 20 years.
Modern pumpsModern pumps are small and lightweightcompared with the early ones (Figures 2 and 3).The pumps are battery operated and hold enoughinsulin for several days, depending on daily need.The infusion rate can be programmed for bothdose and timing. Different basal rates can bepreset; for example, overnight could be lower thanduring the day or vice versa. Bolus boosts can begiven starting just before meals (if analogueinsulins are used), and infusion rates can bereduced during exercise.
The newer pumps are more reliable,38 and mayhave alarms for empty cartridges, low batteries,
Background
8
FIGURE 2 Disetronic H-TRON
FIGURE 3 Medtronic MiniMed 508
occlusion of tubing and faulty electronics, givingrise to less fear of undetected malfunction, whichwas a problem with some of the older pumps.There were isolated reports of pumps deliveringtoo much insulin (pump ‘run-away’, reported withthe Graseby MS3645) or being susceptible toelectrical forces.46 Mobile phones were thought tobe capable of upsetting pump function and thisappeared to be true with at least one older pump(the Microjet Quark) and one older phone, butonly when the phone was transmitting atmaximum power (8 W) and placed in directcontact with the pump.47 No such problem wasreported with the MiniMed pump regardless ofmodel of phone and direct contact. Modernphones have much lower maximum power, of1–2 W,47 and modern pumps have mechanisms toprotect against over-delivery.48
Why are pumps little used in the UK?Despite the invention of the technique in the UK,pumps are used much less there than in other
countries such as the USA, Germany, France andItaly (Table 3, Figure 4). The reasons for thisprobably include conservatism and cost concerns,but there have also been concerns about earlyreports of episodes of diabetic ketoacidosis (DKA)on CSII. A series of papers in the mid-1980sreported adverse events. Mecklenburg andcolleagues in 198449 found that diabetic controlimproved considerably on pump therapy but that42% of 161 patients experienced one or moreadverse events, the most serious being ketoacidosis(38 episodes in 26 patients; once in every 78patient months on CSII), which was commonerafter CSII than before, and commoner than in acomparison group on conventional insulininjections. The comparison group was notobtained by randomisation but some measureswere taken to make it reasonably well matched,except for pregnancies, with the pump group.Patients suffering DKA fell into three groups. Thelargest number had no DKA in the 5–25 monthsbefore starting pump therapy. The two othergroups consisted of those who had DKA bothbefore and after (pumps made little difference)and those who had DKA before but not after. Most
Health Technology Assessment 2004; Vol. 8: No. 43
9
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
TABLE 3 Estimated pump use
Country Numbers Funding
USA 140,000 Fully funded by Medicare and private insuranceGermany 33,000 Fully reimbursed by insurance after 3-month trialFrance 9,500 Fully fundedItaly 5,700 Limited funding in 8 of 25 regionsThe Netherlands 4,700 Fully reimbursed by insurance after 1-month trialSweden 4,200 Fully fundedIsrael 1,900 Fully fundedJapan 1,500 No reliable informationCzech Republic 1,350 Fully fundedSwitzerland 1,300 Rental schemeUK 1,200 Random funding; no cohesive policyNorway 1,000 Fully fundedAustria 1,000 Fully funded by various schemesFinland 750 70% of cost refundedBelgium 750 Diabetic patients receive income support for 70% of costAustralia 500 Private insurance for pumps onlySpain 350 No fundingDenmark 350 Selective funding by hospital budgetsPoland 250 No funding, special cases onlyChina 250 Funded by hospital budgetsIrish Republic 100 Fully fundedSlovenia 50 Full funding for childrenHungary 20 No reliable informationGreece 20 No reliable informationNorthern Ireland 20 Health Boards looking at uniform policy and criteriaLatvia 20 No reliable informationRussia 10 No reliable informationKorea 20,000 No reliable information (pumps produced locally)
Source: INPUT, May 2002, reproduced with permission.
(128 of 161) patients did not have DKA on pumptherapy. The main cause of DKA was intercurrentillness; in half the episodes, patients had not beenchecking urine for ketones, which would havealerted them to the need to increase insulindosage; in half the episodes they had not followedinstructions about pump settings. Most of theepisodes of DKA happened in the first 5 monthsof CSII. Hence the failure may be partlyattributable to lack of education, or perhaps toomuch reliance on pumps, rather than CSII itself,although the authors reported pump malfunction,usually tubing problems, of some kind in 79% ofpatients. The commonest adverse event wasinjection site infection (once every 27 patientmonths)
Also in 1984, Teutsch and colleagues50 reportedthat there had been 35 deaths amongst theestimated 3500 pumps users in the USA at thattime, a rate no different from that expectedamongst non-CSII users. Two deaths appeared tobe directly attributable to pump therapy, onebeing due to pump malfunction and the other tobacterial endocarditis secondary to an infectedinjection site.
In the UK, Knight and colleagues51 in Sheffieldreported 13 episodes of DKA in 150 patient yearsof CSII. Nine of these occurred in the first year ofCSII. There were no instances of mechanicalfailure of the pump itself, but one episodefollowed disconnection of the cannula and anotherwhen the patient did not realise that there was noinsulin left. As in the Mecklenburg study, manyoccurrences of DKA followed other illnesses.
Knight and colleagues concluded that: “At present,the potential problem of severe hyperkalaemiaassociated with ketoacidosis during treatment witha pump creates doubt about the feasibility of CSII … .”
Also in 1984, the Kroc group reported anincreased risk of DKA in CSII compared withconventional therapy, but this time from arandomised study.52 However, the duration wasonly 8 months; eight patients had DKA.
Another adverse report came from Steel andWest,53 in the form of a case history of a pregnant diabetic woman whose baby diedin utero. This was related to incorrect use of thepump.
In 1986, the Sheffield group54 compared patientson CSII who had had DKA with those on CSIIwho had not. The main differences found werelength of full-time education and psychologicalmeasures such as locus of control. Those whosuffered DKA had left education 3 years earlier on average and were less likely to feel that theycould take control of their diabetes. The authors therefore concluded that it may bepossible to select out patients unsuitable for CSII or to provide better education and training.
In the DCCT, by contrast, there was no evidenceof increased DKA risk in pump users and this hasbeen the experience of groups with extensiveexperience of the use of CSII in the UK,55
Germany56 and the USA.57
Background
10
0
1
2
3
4
5
6
7
USA Germany Netherlands Israel France UK
Pum
p us
ers
(% T
ype
1 su
bjec
ts)
FIGURE 4 CSII use in different countries. Source: Pickup J, Guy’s, King’s and St Thomas’ School of Medicine, London: personalcommunication, 2002.
There have been two reports of CSII injection sitesbecoming infected with unusual organisms such asMycobacterium fortuitum58,59 and one set of authorswondered if the risk of infection with organismsfound in soil and water might be increased becauseusers could bathe and swim with some devices.59
The earlier reports nevertheless seem to have givenrise to a widespread belief in the UK that CSII isunsafe. For example, one hospital replying to apatient enquiry about funding a pump, said that:“Problems which arose from use of earlier CSIIsystems include reports of death due to diabeticketoacidosis following accidental disconnection ofthe pump and abscess formation at the site.”
One of the submissions received from patients forthis review said that doctors had old-fashionedopinions about pumps: “Many consultants I havespoken to have preconceived views on the pump.Because pumps were unreliable a few years agothey don’t seem to realise that technology moveson. Look at mobile phones 10 years ago. Look atthem now.”
There could be several reasons why the UK has hada much lower use of pumps than other countries:
� Experience with early, less reliable technologyand the ‘learning curve’
� ‘Migration’ of organised CSII research anddevelopment for want of resources in the UK(suggested by Keen H, Guy’s Hospital: personalcommunication, 2002).
� The fear of ketoacidosis.� Non-prescribability of pumps and adnexae
(perhaps partly based on the fear that largenumbers of patients with Type 1 diabetes wouldbe referred for CSII).
� Cost concerns and competition for fundingfrom other desirable developments. There maybe some clinics where there is support inprinciple for pump use, but where their priorityis lower than other items.
� Manpower resources, especially diabetesspecialist nurses (DSNs).
There is considerable interest in pumps amongstpatients and a growing interest from healthprofessionals, and two new groups have beenformed in the UK. The first is a users’ group,INPUT (Insulin Pump Therapy), which is: “… apatient led support group for diabetics usinginsulin pumps, run by pump users and theirfamilies, an information centre for people seekingfacts about insulin pumps, their use and how toobtain and fund them.”
The main aim of INPUT is to “increase theawareness and understanding of pump therapyand to have it funded by the Department ofHealth” (INPUT, October 2001,www.webshowcase.net/input).
The other group is an educational initiative forprofessionals in diabetes care, Pump Managementfor Professionals (PUMP),60 which organisestraining courses for medical and nursing staff.
Subsidiary questionsChildren, including overnight useGood control is difficult in children, and only aminority achieve it.30 Case series carried out in the1980s showed mixed results.61–64 A review in 1986summarised the finding of the early studies, andnoted conflicting results.65
Children at different ages need to be consideredseparately, perhaps grouped as follows;
� infants and toddlers under constant supervisionby parents
� young children attending nurseries or school,away from parents for part of the day
� older children who might be able to managepumps at school without supervision
� children old enough to look after themselves.
There has been debate about ages at whichchildren can manage their own pumps. It will varyamongst children according to their abilities.
One option that has been used, probablyinfrequently, is for children to have CSII onlyovernight,66,67 when they are under parentalsupervision. Overnight-only CSII has also beenused in adults.68,69
PregnancyCSII has been suggested as a way of normalisingdiabetic control in women with Type 1 diabetesduring pregnancy,70 or preconceptionally in orderto reduce the increased malformation rate seen inpoorly controlled diabetes.71 Various small earlystudies either had no control group72 or involvednon-randomised comparisons.73,74
In a 2000 review, Gabbe70 noted that CSII couldbe particularly useful in pregnancy because thebolus from pumps can be modified to fit with theslower absorption of nutrients associated with thedelayed gastric emptying seen in pregnancy.
Health Technology Assessment 2004; Vol. 8: No. 43
11
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
However, he also noted that there is relativeinsulin resistance in the third trimester ofpregnancy, and that any interruption to insulinsupply could lead to an increased risk of DKA. Herecommended blood glucose testing at 2–3 a.m. inorder to check on insulin delivery.
In Auckland,75 CSII has been used for improvingcontrol in women with Type 2 diabetes andgestational diabetes, particularly those whorequired large (over 100 units per injection) dosesof insulin. These women were mainly Maori orSouth Sea Islanders. Insulin dosage and weightgain were greater on CSII but the comparison wasnot based on an RCT and numbers were small.
The key question for this review is whether CSIIhas any advantages over MDI when usedpreconceptually (for reducing malformations) orduring pregnancy (for improving outcomes for thebaby).
Insulin-resistant Type 2 diabetesA few studies have suggested that short periods ofCSII may be useful in patients with true Type 2diabetes who are poorly controlled on oral drugs.The rationale seems to be that improved controlwill reduce insulin resistance.
Analogue insulins in CSIIThere are theoretical advantages of short-actinganalogues in CSII because of the more rapid effectwhen used for bolus doses. The bolus could begiven when patients sit down to eat, rather than30 minutes before as with soluble, which may beuseful when the time of a meal is uncertain.
One implication is that in comparisons of MDIwith CSII, it is necessary to make sure that bothforms of intensive treatment use the same type ofinsulin. To compare CSII using analogues withMDI using soluble would introduce a confoundingfactor.
Background
12
MethodsThe a priori methods used for systematicallyreviewing evidence of clinical effectiveness weredescribed in the research protocol (Appendix 1),which was sent for expert comments to membersof the advisory panel for the review. Althoughmany helpful comments were received relating tothe general content of the research protocol, therewere none that identified specific problems withthe methods of review. Some changes, additions orpoints of clarification were made to the methodsdiscussed in the original protocol, and these areoutlined below.
� The evidence from the eligible RCTs andcrossover studies was supplemented withinformation from selected non-randomised andobservational studies, to provide information onareas where no evidence eligible for inclusionwas identified. A sensitive search forobservational studies was undertaken. Thesewere filtered by an information scientist andstudies for inclusion were then selected by anexperienced reviewer, looking specifically forstudies that could help answer questions 2, 3, 4and 6 on p. 1. An assessment of quality was notundertaken for these selected studies.
� Importantly it should be noted that the sections‘Quantity and quality of research’ (p. 14), ‘Resultsin adults with Type 1 diabetes’ (p. 16), ‘Pregnancy’(p. 32), ‘Adolescents’ (p. 36) and ‘Analogue versussoluble insulin in CSII’ (p. 38) are based on asystematic review of the evidence from eligibleRCTs and crossover studies, while informationfrom selected observational studies is discussed inthe sections ‘CSII in children’ (p. 37), ‘Overnight-only CSII in adults’ (p. 37), ‘Short-term CSII inpoorly controlled adults with Type 2 diabetes’ (p.37) and ‘Discontinuation rates’ (p. 42).
The methods outlined in the protocol aresummarised below.
Sources of information, search terms and a flowchart outlining the identification of studies aredescribed in Appendix 2. It was decided to run avery sensitive search in order to capture not onlyRCTs for efficacy, but also a wide range ofevidence, including:
� QoL studies� safety and side-effect studies, including a
selection of individual case reports that mightillustrate problems with pumps
� ‘real-life’ case series, especially those that gavedata on discontinuation rates
� a selection of reviews, old and new, partly tolook for reasons why the enthusiasm for pumpswaxed and waned in the last period of pumpactivity in the 1980s
� cost studies� studies on different types of insulin, such as
analogue versus soluble� studies using CSII for short periods to improve
insulin sensitivity in Type 2 diabetes� studies using CSII only for part of the day,
usually at night.
Background papers on the effect of improvedcontrol on long-term complications, mainly fromthe DCCT, Oslo, Stockholm and UKPDS studies,were also obtained. The search was carried out bya skilled information scientist. Abstracts (or justtitles in the minority of studies where no abstractwas available) were checked by two people. Of the3760 studies retrieved, 106 were selected forfurther scrutiny. Some of these were recentreviews, to give a check on completeness ofascertainment of RCTs and other studies. Otherswere exchanges of correspondence oncontroversial areas. A search of the electronic BMJwas carried out to capture correspondencefollowing the 2001 editorial,38 and the referencescited were checked. The Cochrane Library wassearched for RCTs, reviews and protocols and theCochrane Metabolic and Endocrine Group wasconsulted regarding any ongoing activity. Expertsin the field, including some of the pioneers ofpump use, were consulted as part of the advisorypanel. An unpublished meta-analysis andhistorical review were thus obtained.Manufacturers’ submissions to NICE werereviewed as a check on completeness of retrieval ofpublished trials.
The full text of relevant studies was examined andinclusion criteria were applied independently bytwo reviewers. An Access database was designedfor data extraction, which was carried out by onereviewer and checked by a second reviewer. The
Health Technology Assessment 2004; Vol. 8: No. 43
13
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Chapter 3
Clinical effectiveness
quality of eligible RCTs and crossover studies was assessed in accordance with chapter II.5 ofCRD Report 4 (2nd Edition) (Appendix 3).Quality criteria were applied by one reviewer and checked by a second reviewer. At each stage, any differences in opinion were resolvedthrough discussion or consultation with a thirdreviewer.
Inclusion criteriaStudies comparing CSII using insulin pumps withoptimised MDI (at least three injections per day)were included in the review. Comparisons ofanalogue and soluble insulin in CSII were alsoincluded. Comparisons of conventional therapy,implantable pumps and hospital inpatient pumpswere excluded.
The review includes people with insulin-treateddiabetes (Type 1 or Type 2), including adults,children, adolescents and pregnant women. Newlydiagnosed patients were excluded.
Parallel RCTs and randomised and non-randomised crossover studies were included in thereview. Studies with a duration of less than10 weeks on each treatment were excluded.
Studies were included if they reported one ormore of the following as primary outcomes:glycated haemoglobin, insulin dose, weightchange, lipid levels, patient preference, QoL oradverse effects.
Studies in non-English language or available only as abstracts were excluded from the mainanalysis.
Studies excluded from the review are listed inAppendix 4 and a list of recent studies reportedonly as abstracts is given in Appendix 5.
Data synthesisData were synthesised through a narrative reviewwith tabulation of all eligible studies. Data onglycated haemoglobin and insulin dose werecombined where appropriate in a meta-analysissummarising the weighted mean difference(WMD) using the random-effects model. Inassessing glycated haemoglobin, HbA1 and HbA1c
were considered sufficiently similar on clinicalgrounds to be combined within the meta-analysis,focusing on changes rather than absolute values.Further stratification by assay type was notundertaken owing to the lack of additional clarityprovided by the small number of studies reportingat each length of follow-up. A sensitivity analysis
was performed excluding non-randomisedcrossover studies.
CSII versus MDIQuantity and quality of researchTwenty studies comparing CSII and MDI met theinclusion criteria for the review and are shown inTables 4–15 and Appendices 6–13. These includedeight parallel RCTs, nine randomised crossoverstudies and three non-random crossover studies.Fourteen studies included adults with Type 1diabetes, four studies included pregnant women,and two studies included adolescents. One studywas reported in three publications.76–78 Only twoof the studies comparing CSII and MDI usedanalogue insulins for both groups.79,80
No RCTs or crossover studies were identified inchildren [see the section ‘CSII in children’ (p. 37)],CSII overnight only in adults [see the section‘Overnight-only CSII in adults’ (p. 37)] or inpoorly controlled adults with Type 2 diabetes [seethe section ‘Short-term CSII in poorly controlledadults with Type 2 diabetes’ (p. 37)]. Selectedobservational studies were discussed in thesesections, and no assessment of quality wasundertaken.
Adults with Type 1 diabetesThe quality of reporting and methodology of theincluded studies was generally poor by today’sstandards (Table 4). Of the 14 studies includingadults with Type 1 diabetes, the method ofrandomisation was adequate in just two, byBrinchmann-Hansen and colleagues76 and Tsuiand colleagues,79 both of which were parallelRCTs. The three non-randomised crossoverstudies were assessed as having an inadequatemethod of randomisation81–83 and the method wasnot reported in the remaining studies. Allocationconcealment was inadequate in four studies79,81–83
and unknown in 10 studies; therefore, they may besubject to selection bias. Moreover, the similarityof groups at baseline was reported by just fourstudies.76,79,84,85 Eligibility criteria for entry intothe study were reported in eightstudies76,79–81,83,84,86,87 but were not reported insix.82,85,88–91 Point estimates and measure ofvariability were presented for the primary outcomemeasure in most studies, but this was only partiallyreported in three studies where the data werepresented in figures only and had to beextrapolated.76,84,90 Data were analysed accordingto intention-to-treat (ITT) principles in fourstudies79,84–86 and withdrawals were adequately
Clinical effectiveness
14
reported in seven studies.76,80,83,85–87,90 Threestudies79,84,91 partially reported loss to follow-up(numbers but not reasons, or vice versa) and inone study88 the numbers were not specified foreither group. Three studies did not report loss tofollow-up.81,82,89
PregnancyTwo of the four parallel RCTs comparing CSII andMDI in pregnancy were thought to haverandomisation that was partially adequate,92,93 butallocation concealment was inadequate (Table 4).The method of randomisation was not stated inthe studies by Carta and colleagues94 and Burkartand colleagues,95 and therefore concealment ofallocation was unknown. The similarity of groupsat baseline was reported by just two of thestudies.93,94 Eligibility criteria for entry into thestudy were not clearly stated in three studies.92–94
Point estimates and measure of variability werepresented for the primary outcome measure in twostudies;92,93 however, this was only partiallyreported in the study by Carta and colleagues,94
where the data were presented in figures only andhad to be extrapolated, and by Burkart andcolleagues,95 where mean blood glucose or
glycated haemoglobin was not reported. ITTanalysis was not used in the studies and, althoughit may be assumed that no loss to follow-upoccurred, this was not clearly stated in any of thestudies.
AdolescentsThe two studies comparing CSII and MDI96 orCSII, MDI and CSII during the night plus MDIduring the day97 in adolescents were of poormethodological quality (Table 4). The crossoverstudies were said to be randomised, but themethod of randomisation was not specified andallocation concealment was unclear.96,97 Thesimilarity of groups at baseline was not reportedby either study. Eligibility criteria were notreported in the study by Schiffrin and colleagues97
and point estimates and measure of variabilitywere only partially reported in that study as thedata were presented in figures and had to beextrapolated.97 ITT analysis was not used in eitherstudy and, although Schiffrin and colleaguesreported the numbers withdrawn from eachgroup, the reasons for withdrawal were not given97
and Tamborlane and colleagues did not reportloss to follow-up.96
Health Technology Assessment 2004; Vol. 8: No. 43
15
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
TABLE 4 Quality assessment of studies comparing CSII versus MDI
Study Random Allocation Group Eligibility Point ITT Withdrawalsconcealment similarity estimates
Type 1 diabetesBak, 198787 Unknown Unknown Unknown Adequate Adequate Inadequate AdequateBode, 199681 Inadequate Inadequate Unknown Adequate Adequate Inadequate UnknownBrinchmann-Hansen, 198876 Adequate Unknown Reported Adequate Partial Inadequate AdequateChiasson, 198482 Inadequate Inadequate Unknown Unknown Adequate Inadequate UnknownHaakens, 199083 Inadequate Inadequate Unknown Adequate Adequate Inadequate AdequateHanaire-Broutin, 200080 Unknown Unknown Unknown Adequate Adequate Inadequate AdequateHome, 198288 Unknown Unknown Unknown Unknown Adequate Inadequate InadequateNathan, 198289 Unknown Unknown Unknown Unknown Adequate Inadequate UnknownNosadini, 198885 Unknown Unknown Reported Unknown Adequate Adequate AdequateSaurbrey, 198890 Unknown Unknown Unknown Unknown Partial Inadequate AdequateSchiffrin, 198291 Unknown Unknown Unknown Unknown Adequate Inadequate PartialSchmitz, 198986 Unknown Unknown Unknown Adequate Adequate Adequate AdequateTsui, 200179 Adequate Inadequate Reported Adequate Adequate Adequate PartialZiegler, 199084 Unknown Unknown Reported Adequate Partial Adequate Partial
PregnancyBurkart, 198895 Unknown Unknown Unknown Adequate Partial Inadequate UnknownCarta, 198694 Unknown Unknown Reported Unknown Partial Inadequate UnknownCoustan, 198693 Partial Inadequate Reported Unknown Adequate Inadequate UnknownNosari, 199392 Partial Inadequate Unknown Unknown Adequate Inadequate Unknown
AdolescentsSchiffrin, 198497 Unknown Unknown Unknown Unknown Partial Inadequate PartialTamborlane, 198996 Unknown Unknown Unknown Adequate Adequate Inadequate Unknown
See Appendix 3 for description of coding.
Results in adults with Type 1 diabetesFour randomised parallel,76,79,84,85 sevenrandomised crossover80,86–91 and three non-randomised crossover81–83 studies compared CSIIand MDI in adults with Type 1 diabetes, and areshown in Tables 4–10 and Appendices 6 and 10.Two of the 14 trials (one randomised crossover80
and one randomised parallel79) compared CSIIand MDI using the analogue insulin lispro,whereas the remaining studies compared CSII andMDI using regular insulin. Not all studies used theHbA1c that is standard today, but it is thedifference rather than the absolute values thatmatters. Six of the studies report HbA1,
76,82–84,88,91
seven report HbA1c79–81,85,86,89,90 and one reports
blood glucose only.87 Data are discussed in thisreview according to the reported length of follow-up. One caveat is that many of these studies arefairly old and used obsolete technologies.
Glycated haemoglobin and blood glucoseEight studies reported glycated haemoglobin at10 weeks to 4 months of follow-up(Table 5).76,78–80,82,88–91 A mean reduction in
glycated haemoglobin from baseline of between0.5% and 5.1% was found for CSII treatment, andbetween –1.0% to 4.8% for MDI treatment. Muchof this variation may be explained by differencesin baseline glycated haemoglobin between studies,which ranges from 7.7% for the CSII group in theparallel RCT by Tsui and colleagues79 to 13.2% inthe random crossover study by Schiffrin andcolleagues.91 At 10 weeks to 4 months follow-up,five studies found that glycated haemoglobin waslower with CSII than MDI, reporting a meanreduction of between 0.3% and 2.48%.79,80,88,89,91
This difference was statistically significant in justtwo of the studies (0.35%, p < 0.0001;80 1.7%,p = 0.02688), two other studies reported a non-significant reduction79,91 and one study89 did notreport statistical significance. Saurbrey andcolleagues found no difference in glycatedhaemoglobin between CSII and MDI at 10 weeksof follow-up,90 and two trials reported slightlyhigher glycated haemoglobin with CSII comparedwith MDI [8.9% versus 8.7%, p not reported;76,78
9.1% versus 8.7%, p = not statistically significant(ns)82]. Five of the eight studies provided data
Clinical effectiveness
16
TABLE 5 Glycated haemoglobin in adults with Type 1 diabetes
Study Mean glycated haemoglobin (SD) Difference
CSII MDI(MDI – CSII)
Bak, 198787 Glycated haemoglobin not reportedRandom crossover Blood glucose: Blood glucose:N: 20; end of study: 16 Month 6: 7.7 mmol/l (0.7) Month 6: 7.9 mmol/l (0.7), p = ns 0.2 mmol/lBaseline BG not reported
Bode, 199681 Month 12: 7.4% (1.2) 7.7% (1.5), p = ns 0.3%Non-random crossover Month 24: 7.7% (1.7) Not reportedN: 55. Month 36: 7.4% (1.7) Not reportedHbA1c data collected Month 48: 7.4% (1.2) Not reportedBaseline not reported
Brinchmann-Hansen, 198876,78 Month 3: 8.9% 8.7% –0.2%Parallel RCT Month 6: 9.2% 8.8% –0.4%CSII n: 15; MDI n: 15 Month 12: 8.5% 8.5% 0%HbA1 data collected Month 24: 8.7% (1.16) 9.1% (1.55) 0.4%Baseline CSII: 10.1% (1.55) Month 41: 9.1% 9.4% 0.3%Baseline MDI: 9.4% (1.55)
Chiasson, 198482 Month 3: 9.1% (1.04) 8.7% (1.39), p = ns –0.4%Non-random crossoverN: 12HbA1 data collectedBaseline: 11.9% (2.08)
Haakens, 199083 Month 6: 9.6% (2.47) 9.8% (1.85), p = ns 0.2%Non-random crossoverN: 52; end of study: 35HbA1 data collectedBaseline: 10.4% (1.85)
continued
Health Technology Assessment 2004; Vol. 8: No. 43
17
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
TABLE 5 Glycated haemoglobin in adults with Type 1 diabetes (cont’d)
Study Mean glycated haemoglobin (SD) Difference
CSII MDI(MDI – CSII)
Hanaire-Broutin, 200080 Month 4: 7.89% (0.77) 8.24% (0.77), p<0.001 0.35%Random crossoverN: 41; end of study: 40HbA1c data collectedBaseline: 8.39% (0.87)
Home, 198288 Month 2.5: 10% (2.2) 11.7% (1.9), p = 0.026 1.70%Random crossoverN: 11; end of study: 10HbA1 data collectedBaseline: 10.7% (1.9)
Nathan, 198289 Month 3: 5.4% (0.34) 7.88% (1.37), p not reported 2.48%Random crossoverN: 5HbA1c data collectedBaseline: 9.03% (1.43)
Nosadini, 198885 CSII fixed night basal rate: MDI: 0.8% (MDI – Parallel RCT Month 12: 6.3% (0.7) 7.1% (0.9), p <0.01 (vs CSII fixed) CSII fixed)CSII fixed n: 19; MDI n: 15CSII variable n: 10 CSII with variable night basal rate: 1% (MDI – CSII HbA1c data collected Month 12: 6.1% (0.9), p < 0.01 variable)Baseline not reported (vs MDI)
Saurbrey, 198890 Month 2.5: 7.5% 7.5% 0%Random crossoverN: 21, end of study: 19HbA1c data collectedBaseline (CSII then MDI): 8.7% (1.74)Baseline (MDI then CSII) 8.8% (2.17)
Schiffrin, 198291 Month 3: 8.1% (0.6) 8.4% (0.7), p = ns 0.3%Random crossover. Month 6: 8.2% (0.5) 8.4% (0.5), p = ns 0.2%N: 20; end of study: 16HbA1 data collectedBaseline: 13.2% (1.1)
Schmitz, 198986 Month 6: 7% (1) 7.7% (1), p = 0.002 0.7%Random crossoverN: 10HbA1c data collectedBaseline: 7.6% (0.9)
Tsui, 200179 Month 3: 6.92% 7.55%. Treatment effect (adjusted 0.63%Parallel RCT for baseline HbA1c) = –0.21 CSII N:13; end of study: 12 (95% CI –0.59 to 0.17), p > 0.10MDI N: 14; end of study:14 Month 6: 7.19% 7.62%. Treatment effect (adjusted 0.43%HbA1c data collected for baseline HbA1c) = –0.01 Baseline CSII: 7.73% (0.6) (95% CI –0.44 to 0.42), p > 0.10Baseline MDI: 8.16% (0.7) Month 9: 7.38% 7.56%. Treatment effect (adjusted 0.18%
for baseline HbA1c) = 0.25 (95% CI –0.19, 0.68), p >0.10
Ziegler, 199084 Month 6: 8.2% 8.8%, p < 0.05 0.6%Parallel RCT Month 12: 8.5% 8.7%, p = ns 0.2%CSII N: 49; end of study: 36 Month 18: 8.5% 8.4%, p = ns –0.1%MDI N: 47; end of study: 37 Month 24: 8.6% 8.5%, p = ns –0.1%HbA1 data collectedBaseline CSII: 9.8%Baseline MDI: 9.5%, p = ns
BG, blood glucose; N, number of patients.
[means and standard deviations (SDs)] thatallowed them to be combined in a meta-analysis(Figure 5).80,82,88,89,91 Pooling the data using arandom-effects model (�2 test for heterogeneity15.68, df = 4, p = 0.0035) showed an overallreduction in glycated haemoglobin with CSIIcompared with MDI (–0.64, 95% CI –1.28 to 0.01).Because of the heterogeneity, these results shouldbe viewed with some caution, but theheterogeneity was around the effect size. Thestudy by Home and colleagues88 used a sub-optimal MDI regime with bovine ultralenteinsulin, and HbA1c actually worsened during MDIcompared with conventional treatment at baseline,thereby exaggerating the effects of CSII. Hencethere is an argument for excluding this study onthe grounds that the control arm did not receiveoptimal MDI. Repeating the meta-analysis without
this study (Figure 6) reduced the effect size (–0.52,95% CI –1.19 to 0.14, p = 0.12).
Six studies reported glycated haemoglobin at6 months follow-up (Table 5).76,78,79,83,84,86,91 Amean reduction from baseline of between 0.54%and 5% was found for CSII and between –0.1 and4.8% for MDI. Five of the studies found thatglycated haemoglobin was lower with CSII thanMDI, reporting a mean reduction of between 0.7%and 0.2%.79,83,84,86,91 This difference wasstatistically significant in just two of the studies(0.7%, p = 0.00286; 0.6%, p < 0.0584).Brinchmann-Hansen and colleagues found slightlyhigher glycated haemoglobin with CSII than withMDI (9.2% versus 8.8%), but significance was notreported.76,78 Three of the six studies reportingglycated haemoglobin at 6 months could be
Clinical effectiveness
18
01 2.5 to 4 monthsx Brinchmann, 1988
Chiasson, 1984Hanaire-Broutin, 2000Home, 1982Nathan, 1982
x Saurbrey, 1988Schiffrin, 1982
x Tsui, 2001Subtotal (95% CI)Test for heterogeneity χ2 = 15.68, df = 4, p = 0.0035Test for overall effect z = 1.93, p = 0.05
02 6 monthsx Brinchmann, 1988
Haakens, 1990Schiffrin, 1982Schmitz, 1989
x Tsui, 2001x Ziegler, 1990Subtotal (95% CI)Test for heterogeneity χ2 = 1.10, df = 2, p = 0.58Test for overall effect z = 1.67, p = 0.10
04 12 monthsBode, 1996
x Brinchmann, 1988Nosadini, 1988
x Ziegler, 1990Subtotal (95% CI)Test for heterogeneity χ2 = 2.57, df = 1, p = 0.11Test for overall effect z = 1.75, p = 0.08
15 8.90 (0.00) 14 8.70 (0.00) 12 9.10 (1.04) 12 8.70 (1.39) 40 7.89 (0.77) 40 8.24 (0.77) 10 10.00 (2.20) 10 11.70 (1.90) 5 5.40 (0.34) 5 7.88 (1.37) 19 7.50 (0.00) 19 7.50 (0.00) 16 8.10 (0.60) 16 8.40 (0.70) 12 6.92 (0.00) 14 7.55 (0.00)129 130
15 9.20 (0.00) 14 8.80 (0.00) 35 9.60 (2.47) 35 9.80 (1.85) 16 8.20 (0.50) 16 8.40 (0.50) 10 7.00 (1.00) 10 7.70 (1.00) 12 7.19 (0.00) 14 7.62 (0.00) 36 8.20 (0.00) 37 8.80 (0.00)124 126
55 7.40 (1.20) 55 7.70 (1.50) 15 8.50 (0.00) 14 8.50 (0.00) 19 6.10 (0.90) 10 7.10 (0.90) 36 8.50 (0.00) 37 8.70 (0.00)125 116
0.0 Not estimable18.5 0.40 (–0.58 to 1.38)29.7 –0.35 (–0.69 to –0.01)
9.2 –1.70 (–3.50 to 0.10)14.7 –2.48 (–3.72 to –1.24)
0.0 Not estimable27.9 –0.30 (–0.75 to –0.15)
0.0 Not estimable100.0 –0.64 (–1.28 to 0.01)
0.0 Not estimable9.0 –0.20 (–1.22 to 0.82)
78.7 –0.20 (–0.55 to 0.15)12.3 –0.70 (–1.58 to 0.18)
0.0 Not estimable0.0 Not estimable
100.0 –0.26 (–0.57 to 0.05)
55.8 –0.30 (–0.81 to 0.21)0.0 Not estimable
44.2 –1.00 (–1.69 to –0.31)0.0 Not estimable
100.0 –0.61 (–1.29 to 0.07)
–4 –2 2 40Favours CSII Favours MDI
Study mean (SD)CSII
n mean (SD)MDI
nWMD
(95% CI Random)Weight
%WMD(95% CI Random)
FIGURE 5 Meta-analysis of the effect of CSII versus MDI on glycated haemoglobin in adults with Type 1 diabetes. Note: HbA1 isreported by Brinchmann-Hansen, 1988,76 Chiasson, 1984,82 Haakans, 1990,83 Home, 1982,88 Schiffrin, 198291 and Ziegler, 1990.84
HbA1c is reported by Bode, 1996,81 Hanaire-Broutin, 2000,80 Nathan, 1982,89 Nosadini, 1988,85 Saurbrey, 1988,90 Schmitz, 198986
and Tsui, 2001.79
combined in a meta-analysis (Figure 5).83,86,91
Pooling the data using a random-effects model (�2
test for heterogeneity = 1.10, df = 2, p = 0.58)showed a non-significant reduction in glycatedhaemoglobin with CSII compared with MDI(–0.26, 95% CI –0.57 to 0.05).
Only one study reported glycated haemoglobin at9 months of follow-up79 (Table 5). Tsui andcolleagues found a reduction from baseline of0.35% with CSII and 0.60% with MDI. A non-significant difference between CSII and MDI wasfound (7.38% versus 7.56%) at 9 months of follow-up, with a treatment effect of 0.25 (95% CI –0.19to 0.68, p > 0.10) adjusted for baseline glycatedhaemoglobin.
Four studies reported glycated haemoglobin at12 months follow-up (Table 5).76,78,81,84,85 Only twostudies report baseline data, demonstrating amean reduction from baseline of 1.3%84 to1.6%76,78 for CSII and 0.8%84 to 0.9%76,78 forMDI. Nosadini and colleagues report significantlylower glycated haemoglobin with both fixed rate(6.3%) and variable rate (6.1%) CSII comparedwith MDI (7.1%, p < 0.01).85 However, threestudies report no difference between CSII andMDI (0% to 0.2%, p = ns).76,78,81,84 Two of the fourstudies provided data that allowed them to becombined in a meta-analysis (Figure 5).81,85 As thestudy by Nosadini and colleagues85 had twogroups with CSII, only the group with the variableinsulin rate was included in the meta-analysis, asthis had the lowest mean glycated haemoglobinlevel and may provide optimal control. Pooling thedata using a random-effects model (�2 test forheterogeneity 2.57, df = 1, p = 0.11) revealed anon-significant result in favour of CSII (–0.61,95% CI –1.29 to 0.07).
Two studies reported glycated haemoglobin at24 months of follow-up (Table 5), showing a meanreduction from baseline of between 1.2%84 and1.4%76,78 for CSII and between 0.3%76,78 and 1%84
for MDI. Brinchmann-Hansen and colleaguesreport lower glycated haemoglobin with CSII thanMDI (8.7% versus 9.1%), although significance isnot reported,76,78 and Ziegler and colleaguesreport no significant difference between CSII andMDI (8.6% versus 8.5%, p = ns) at 12 months.84
Studies could not be combined into a meta-analysis, as SDs were not presented.
Bode and colleagues81 reported glycatedhaemoglobin at 24 (7.7%, SD 1.7), 36 (7.4%, SD1.7) and 41 months (7.4%, SD 1.2) follow-up forCSII treatment only (Table 5). This was a
prospective non-randomised crossover study inwhich patients were switched to CSII after at least12 months with MDI.
Brinchmann-Hansen and colleagues76,78 reportedglycated haemoglobin at 41 months of follow-up(Table 5), demonstrating a mean reduction of 1.0% from baseline with CSII and no change withMDI. Glycated haemoglobin was lower with CSIIthan with MDI at 41 months of follow-up (8.7%versus 9.1%), but statistical significance was notreported.
Bak and colleagues87 did not report glycatedhaemoglobin, presenting blood glucose at12 months but not at baseline (Table 5). This studyshows a non-significant improvement in bloodglucose with CSII compared with MDI [7.7(SD 0.7) mmol/l versus 7.9 (SD 90.7) mmol/l, p = ns).
Glycated haemoglobin: sensitivity analysisA further meta-analysis was conducted without thethree non-randomised studies (Chiasson andcolleagues,82 3 months of follow-up; Haakens andcolleagues,83 6 months of follow-up; and Bode andcolleagues,81 12 months of follow-up), as there is agreater possibility of bias in non-randomised trials.
At 10 weeks to 4 months of follow-up, fourrandomised trials could be combined in a meta-analysis (Figure 7).80,88,89,91 Pooling the data usinga random-effects model (�2 test for heterogeneity12.95, df = 3, p = 0.0047) showed a significantreduction in glycated haemoglobin favouring CSIIcompared with MDI (–0.87, 95% CI –1.59 to–0.16). This shows a slightly greater reduction withCSII compared with the earlier meta-analysis(Figure 5). However, even with the non-randomstudies excluded, significant heterogeneity was stillpresent.
At 6 months of follow-up, two randomised trialscould be combined in a meta-analysis(Figure 7).86,91 Pooling the data using a random-effects model (�2 test for heterogeneity 1.08,df = 1, p = 0.3) showed that glycatedhaemoglobin was not significantly reduced withCSII compared with MDI at 6 months of follow-up(–0.28, 95% CI –0.64 to 0.08). This is similar tothe earlier meta-analysis (Figure 5).
Removing the study by Bode and colleagues81
at 12 months of follow-up leaves just onerandomised trial presenting means and standarddeviations; therefore, a meta-analysis cannot becarried out.
Health Technology Assessment 2004; Vol. 8: No. 43
19
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
The shortage of long-term studies is unfortunate,since it is likely that it takes some time foroptimum dosage to be worked out and for patientsto become familiar with adjusting their insulindosages, and perhaps in particular experimentingwith different basal rates at different times of dayand night. Nevertheless, the limited amount ofdata from the trials does suggest that results mayimprove with longer duration.
SummaryThe evidence is consistent in showing animprovement in glycated haemoglobin on CSIIcompared with MDI, the difference averagingabout 0.5%.
CaveatAs has been pointed out to us (Home, PD,Department of Diabetes, The Medical School,
Clinical effectiveness
20
01 2.5 to 4 monthsx Brinchmann, 1988
Chiasson, 1984Hanaire-Broutin, 2000Nathan, 1982
x Saurbrey, 1988Schiffrin, 1982
x Tsui, 2001Subtotal (95% CI)Test for heterogeneity χ2 = 13.64, df = 3, p = 0.0034Test for overall effect z = 1.55, p = 0.12
15 8.90 (0.00) 14 8.70 (0.00) 12 9.10 (1.04) 12 8.70 (1.39) 40 7.89 (0.77) 40 8.24 (0.77) 5 5.40 (0.34) 5 7.88 (1.37) 19 7.50 (0.00) 19 7.50 (0.00) 16 8.10 (0.60) 16 8.40 (0.70) 12 6.92 (0.00) 14 7.55 (0.00)119 120
0.0 Not estimable20.1 0.40 (–0.58 to 1.38)33.0 –0.35 (–0.69 to –0.01)16.0 –2.48 (–3.72 to –1.24)
0.0 Not estimable30.9 –0.30 (–0.75 to 0.15)
0.0 Not estimable100.0 –0.52 (–1.19 to 0.14)
–4 –2 2 40Favours CSII Favours MDI
Study mean (SD)CSII
n mean (SD)MDI
nWMD
(95% CI Random)Weight
%WMD(95% CI Random)
FIGURE 6 Meta-analysis of the effect of CSII versus MDI on glycated haemoglobin in adults with Type 1 diabetes, excluding Homeand colleagues.88 Note: HbA1 is reported by Brinchmann-Hansen, 1988,76 Chiasson, 1984,82 and Schiffrin, 1982.91 HbA1c is reportedby Hanaire-Broutin, 2000,80 Nathan, 1982,89 Saurbrey, 198890 and Tsui, 2001.79
01 2.5 to 4 monthsx Brinchmann, 1988
Hanaire-Broutin, 2000Home, 1982Nathan, 1982
x Saurbrey, 1988Schiffrin, 1982
x Tsui, 2001Subtotal (95% CI)Test for heterogeneity χ2 = 12.95, df = 3, p = 0.0047Test for overall effect z = 2.39, p = 0.02
02 6 monthsx Brinchmann, 1988
Schiffrin, 1982Schmitz, 1989
x Tsui, 2001x Ziegler, 1990Subtotal (95% CI)Test for heterogeneity χ2 = 1.08, df = 1, p = 0.3Test for overall effect z = 1.52, p = 0.13
15 8.90 (0.00) 14 8.70 (0.00) 40 7.89 (0.77) 40 8.24 (0.77) 10 10.00 (2.20) 10 11.70 (1.90) 5 5.40 (0.34) 5 7.88 (1.37) 19 7.50 (0.00) 19 7.50 (0.00) 16 8.10 (0.60) 16 8.40 (0.70) 12 6.92 (0.00) 14 7.55 (0.00)117 118
15 9.20 (0.00) 14 8.80 (0.00) 16 8.20 (0.50) 16 8.40 (0.50) 10 7.00 (1.00) 10 7.70 (1.00) 12 7.19 (0.00) 14 7.62 (0.00) 36 8.20 (0.00) 37 8.80 (0.00) 89 91
0.0 Not estimable36.4 –0.35 (–0.69 to –0.01)11.3 –1.70 (–3.50 to 0.10)18.1 –2.48 (–3.72 to –1.24)
0.0 Not estimable34.2 –0.30 (–0.75 to 0.15)
0.0 Not estimable100.0 –0.87 (–1.59 to –0.16)
0.0 Not estimable83.8 –0.20 (–0.55 to 0.15)16.2 –0.70 (–1.58 to 0.18)
0.0 Not estimable0.0 Not estimable
100.0 –0.28 (–0.64 to 0.08)
–4 –2 2 40Favours CSII Favours MDI
Study mean (SD)CSII
n mean (SD)MDI
nWMD
(95% CI Random)Weight
%WMD(95% CI Random)
FIGURE 7 Meta-analysis of the effect of CSII versus MDI on glycated haemoglobin in adults with Type 1 diabetes, excluding non-random studies. Note: HbA1 is reported by Brinchmann-Hansen, 1988,76 Home, 1982,88 Schiffrin 1982,91 and Ziegler, 1990.84 HbA1cis reported by Hanaire-Broutin, 2000,80 Nathan, 1982,89 Saurbrey, 1988,99 Schmitz, 198986 and Tsui, 2001.79
Newcastle: personal communication, 2002), HbA1c
may not be a perfect guide to diabetes control inpeople who achieve good overall HbA1c levels butwho have problems with hypoglycaemia, since thehypoglycaemic events will reduce the HbA1c. IfCSII reduces the frequency of hypoglycaemia, ofwhatever degree, that will tend to increase HbA1c,and this may conceal the extent of theimprovement in control of hyperglycaemia.
Daily insulin doseInsulin dose was reported in the studies usingeither units per day (U/day) or units per kilogramper day (U/kg/day), and the data are discussedhere for each method separately, according to thereported length of follow-up.
Five studies reported insulin dose at 10 weeks to4 months follow-up (Table 6), with four studiespresenting insulin in units per day80,82,88,89 andone study presenting insulin units per kilogramper day.90 Compared with baseline, a meandecrease in insulin dose of 1.2–9 U/day wasobserved with CSII and a mean increase of3.7–20 U/day was observed with MDI.80,82,88,89
Three studies found significantly less insulin wasrequired with CSII than with MDI(8.80–29 U/day).80,82,88 The fourth study, byNathan and colleagues, found a lower daily insulindose with CSII compared with MDI [35.4 (SD11.5) U/day versus 48.8 (SD 13.18) U/day] but didnot report statistical significance.89 Pooling thesedata into a meta-analysis using a random-effectsmodel (�2 test for heterogeneity 3.78, df = 3,p = 0.29) shows an overall significant reduction ofabout 12 U/day in daily insulin dose after 10 weeksto 4 months of treatment on CSII compared withMDI (–11.90 units, 95% CI –18.16 to –5.63)(Figure 8). Repeating the meta-analysis without thestudy by Home and colleagues,88 which used asuboptimal MDI regime with bovine insulin,(Figure 9) slightly reduces the effect size (–9.73,95% CI –14.55 to –4.91), but the decrease ininsulin dose with CSII remains significant.
Saurbrey and colleagues90 found a mean reductionin insulin dose from baseline of 0.05 and0.03 U/kg/day at 10 weeks of follow-up for CSIIand MDI, respectively. Insulin use was slightlylower with CSII than MDI (0.62 versus0.64 U/kg/day), but statistical significance was notpresented.
Four studies83,86,87,91 reported insulin dose at6 months of follow-up (Table 6); two87,91 presentthe data as units per day and two83,86 units perkilogram per day. Insulin dose increased from
baseline in the study by Bak and colleagues (CSII0.6, MDI 5.9 U/day),87 but decreased in the studyby Schiffrin and colleagues (CSII –6, MDI–4 U/day).91 Haakens and colleagues reportedinsulin dose for MDI with isophane and withultralente, with a difference from baseline of –0.04and 0.03 U/kg/day (p = ns), respectively. Insulindose with CSII was significantly lower than atbaseline (0.64 versus 0.76 U/kg/day, p < 0.001).83
Schmitz and colleagues reported the meandifference from baseline for CSII as 0.05 U/kg/day(p = 0.03) and for MDI as 0 U/kg/day (p = ns).86
Daily insulin dose at 6 months of follow-up waslower with CSII than with MDI: between 2.0(p = ns)91 and 5.30 U/day (p < 0.05),87 andbetween 0.05 (p = 0.02)86 and 0.08 U/kg/day(p < 0.005) for isophane and 0.15 U/kg/day(p < 0.001) for ultralente.83 The differencebetween isophane and ultralente was alsostatistically significant (p < 0.01). A meta-analysisof insulin dose at 6 months of follow-up could not be performed as two studies did not present SDs.
Insulin dose was not often reported after 6 monthsof follow-up (Table 6). At 9 months of follow-up,Tsui and colleagues79 found little differencecompared with baseline (0.1 U/kg/day decrease forCSII and no change for MDI) or betweentreatments (0.1 U/kg/day, p = ns). Bode andcolleagues81 reported a non-significant decrease of6.50 U/day with CSII compared with MDI at12 months of follow-up, but did not reportbaseline values. As with glycated haemoglobin, theauthors also reported insulin dose for CSII at 24(36.6 U/day), 36 (37.7 U/day) and 48 months(37.8 U/day), but did not follow patients on MDIfor this length of time.
At 24 months of follow-up, Brinchmann-Hansenand colleagues78 found a decrease in insulin doseof 0.12 U/kg/day for CSII compared with baselineand no change for MDI (Table 6). There was nosignificant difference between CSII and MDI[0.68 U/kg/day (SD 0.19) versus 0.72 U/kg/day(SD 0.72), p = ns].
Daily insulin dose: sensitivity analysisA further meta-analysis was conducted without thenon-randomised study (Chiasson and colleagues82)at 10 weeks to 4 months follow-up (Figure 10).Pooling the three randomised studies80,88,89 thatreported data in units per day using a random-effects model (�2 test for heterogeneity 3.72, df = 2,p = 0.16) showed that the decrease in insulin dosewith CSII compared with MDI remained significant(WMD –13.63, 95% CI –23.62 to –3.64).
Health Technology Assessment 2004; Vol. 8: No. 43
21
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Clinical effectiveness
22
TABLE 6 Insulin dose in adults with Type 1 diabetes
Study Mean insulin dose (SD) Difference
CSII MDI(MDI – CSII)
Bak, 198787 Month 6: 46.6 U/day (10) 51.9 U/day (12.9), p < 0.05 5.3 U/dayRandom crossoverN: 20; end of study: 16Baseline: 46 U/day (16)
Bode, 199681 Month 12: 36.4 U/day (12.1) 42.9 U/day (17.9), p = ns 6.5 U/dayNon-random crossover Month 24: 39.6 U/day (14.4)N: 55 Month 36: 37.7 U/day (13.1)Baseline not reported Month 48: 37.8 U/day (14.2)
Brinchmann-Hansen, 198878 Month 24: 0.68 U/kg/day (0.19) 0.72 U/kg/day (0.12), p = ns 0.04 U/kg/dayParallel RCTCSII N: 15; MDI N: 15Baseline CSII: 0.8 U/kg/day (0.23)Baseline MDI: 0.72 U/kg/day (0.08)
Chiasson, 198482 Month 3: 43.9 U/day (10) 56.1 U/day (20.4), p < 0.01 12.2 U/dayNon-random crossoverN: 12Baseline: 44.7 U/day (14.6)
Haakens, 199083 Month 6: 0.64 U/kg/day (0.18) Month 6: MDI/human isophane 0.08 U/kg/dayNon-random crossover 0.72 U/kg/day (0.25), p < 0.005N: 52; end of study: 35 Month 6: MDI/human ultralente 0.15 U/kg/dayBaseline: 0.76 U/kg/day (0.25). 0.79 U/kg/day (0.25), p < 0.001
(isophane vs ultralente, p < 0.01)
Hanaire-Broutin, 200080 Month 4: 38.5 U/day (9.8) 47.3 U/day (14.9), p < 0.0001 8.8 U/dayRandom crossoverN: 41; end of study: 40Baseline: 43.6 U/day (13.5)
Home, 198288 Month 2.5: 51 U/day (15.8) 80 U/day (28.5), p = 0.004 29 U/dayRandom crossoverN: 11; end of study: 10Baseline: 60 U/day (18.97)
Nathan, 198289 Month 3: 35.4 U/day (11.5) 48.8 U/day (13.18) 13.4 U/dayRandom crossoverN: 5Baseline: 42.8 U/day (16.04)
Saurbrey, 198890 Month 2.5: 0.62 U/kg/day (0.17) 0.64 U/kg/day (0.13) 0.02 U/kg/dayRandom crossover (values collected during (values collected during N: 21; end of study: 19 treatment period) treatment period)Baseline: 0.67 U/kg/day (0.4)
Schiffrin, 198291 Month 6: 42 U/day 44 U/day, p = ns 2 U/dayRandom crossoverN: 20, end of study: 16Baseline: 48 U/day
Schmitz, 198986 Month 6: 0.55 U/kg/day (0.1) Month 6: 0.6 U/kg/day (0.11), 0.05 U/kg/dayRandom crossover p = 0.02N: 10Baseline: 0.6 U/kg/day (0.11)
Tsui, 200179 Month 9: 0.6 U/kg (0.2) 0.7 U/kg/day (0.2). Treatment 0.1 U/kg/dayParellel RCT effect (adjusted for baseline insulin):CSII N: 13; end of study: 12 –0.10 (95% CI–0.26 to 0.07),MDI N: 14, end of study: 14 p > 0.10Baseline CSII: 0.7 U/kg/day (0.2)Baseline MDI: 0.7 U/kg/day (0.1)
continued
Health Technology Assessment 2004; Vol. 8: No. 43
23
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
TABLE 6 Insulin dose in adults with Type 1 diabetes (cont’d)
Study Mean insulin dose (SD) Difference
CSII MDI(MDI – CSII)
Ziegler, 199084 Values not reported, but it is Parallel RCT stated that daily insulin dose was CSII N: 49; end of study: 36 similar in CSII and MDI at entry MDI N: 47, end of study: 37 and during follow-upBaseline CSII: 49 U/day (17)Baseline MDI: 48 U/day (14)
01 2.5 to 4 months – U/dayChiasson, 1984Hanaire-Broutin, 2000Home, 1982Nathan, 1982
Subtotal (95% CI)Test for heterogeneity χ2 = 3.78, df = 3, p = 0.28Test for overall effect z = 3.72, p = 0.0002
12 43.90 (10.00) 12 56.10 (20.40) 40 38.50 (9.80) 40 47.30 (14.90) 10 51.00 (15.80) 10 80.00 (28.50) 5 35.40 (11.50) 5 48.80 (13.18) 67 67
19.3 –12.20 (–25.05 to 0.65)57.5 – 8.80 (–14.33 to –3.27)
8.8 –29.00 (–49.20 to –8.80)14.4 –13.40 (–28.73 to 1.93)
100.0 –11.90 (–18.16 to –5.63)
Study mean (SD)CSII
n mean (SD)MDI
nWMD
(95% CI Random)Weight
%WMD(95% CI Random)
–100 –50 0 50 100Favours MDIFavours CSII
FIGURE 8 Meta-analysis of the effects of CSII versus MDI on insulin dose (U/day) in adults with Type 1 diabetes
01 2.5 to 4 months – U/dayChiasson, 1984Hanaire-Broutin, 2000Nathan, 1982
Subtotal (95% CI)Test for heterogeneity χ2 = 0.47, df = 2, p = 0.79Test for overall effect z = 3.96, p = 0.00008
12 43.90 (10.00) 12 56.10 (20.40) 40 38.50 (9.80) 40 47.30 (14.90) 5 35.40 (11.50) 5 48.80 (13.18) 57 57
14.1 –12.20 (–25.05 to 0.65)76.1 – 8.80 (–14.33 to –3.27)
9.9 –13.40 (–28.73 to 1.93)100.0 – 9.73 (–14.55 to –4.91)
Study mean (SD)CSII
n mean (SD)MDI
nWMD
(95% CI Random)Weight
%WMD(95% CI Random)
–100 –50 0 50 100Favours MDIFavours CSII
FIGURE 9 Meta-analysis of the effects of CSII versus MDI on insulin dose (U/day) in adults with Type 1 diabetes, excluding Home andcolleagues88
01 2.5 to 4 months – U/dayHanaire-Broutin, 2000Home, 1982Nathan, 1982
Subtotal (95% CI)Test for heterogeneity χ2 = 3.72, df = 2, p = 0.16Test for overall effect z = 2.87, p = 0.008
40 38.50 (9.80) 40 47.30 (14.90) 10 51.00 (15.80) 10 80.00 (28.50) 5 35.40 (11.50) 5 48.80 (13.18) 55 55
56.0 – 8.80 (–14.33 to –3.27)18.0 –29.00 (–49.20 to –8.80)26.1 –13.40 (–28.73 to 1.93)
100.0 –13.63 (–23.62 to –3.64)
Study mean (SD)CSII
n mean (SD)MDI
nWMD
(95% CI Random)Weight
%WMD(95% CI Random)
–100 –50 0 50 100Favours MDIFavours CSII
FIGURE 10 Meta-analysis of the effects of CSII versus MDI on insulin dose in adults with Type 1 diabetes, excluding non-random studies
SummaryIn short-term studies, insulin dose is lower onCSII than on MDI, by about 12 units (20–25%),but there is little difference in longer term studies.
Body weightEight of the studies included in this reviewreported data on change in body weight in adultswith Type 1 diabetes (Table 7).78,80,81,83,84,86–88
Two studies reported weight change at 10 weeks to4 months of follow-up, with a mean increase frombaseline of 0.580 to 1.3 kg88 and –0.688 to 0.8 kg80
for CSII and MDI, respectively. Home andcolleagues88 reported a significantly greater weightin individuals with CSII than with MDI (68.9
versus 67 kg, p = 0.023), whereas Hanaire-Broutinand colleagues80 report a non-significant decreaseof 0.3 kg with CSII.
Three studies reported weight change at 6months of follow-up.83,86,87 Body weight hadincreased from baseline in both groups, by 0.5 to 2.2 kg in CSII and by 0.4 to 2.0 kg inMDI.83,86,87 Haakens and colleagues83 reported a significantly lower weight with CSII comparedwith MDI (69.1 versus 70.6 kg, p < 0.05);conversely, Bak and colleagues87 found that weight was lower with MDI (76.6 versus 74.8 kg, p < 0.01). Schmitz and colleagues86 alsofound weight to be lower with MDI (73 versus72.8 kg), but this difference was not statisticallysignificant.
Clinical effectiveness
24
TABLE 7 Weight change in adults with Type 1 diabetes
Study Mean weight (SD) (kg) Difference
CSII MDI(MDI – CSII) (kg)
Bak, 198787 Month 6: 76.6 (9.6) 74.8 (9.7), p < 0.01 –1.8Random crossoverN: 20; end of study: 16Baseline: 74.4 kg (9.7)
Bode, 199681 Month 12: 68.1 (14.1) 67.4 (14.4), p = ns –0.7Non-random crossover Month 24: 69.7 (15.2)N: 55 Month 36: 70.0 (14.6)Baseline not reported Month 48: 68 (15.2)
Brinchmann-Hansen 198878 Month 24: 70.5 (8.9) 75.1 (11.2), p = ns 4.6Parallel RCTCSII N: 15; MDI N: 15Baseline: 68.6 kg (9.3)
Haakens, 199083 Month 6: 69.1 (8.63) 70.6 (8.63), p < 0.05 1.5Non-random crossoverN: 52; end of study: 35Baseline: 68.6 kg (8.63)
Hanaire-Broutin, 200080 Month 4: 68.7 (10) 69 (9.5), p = ns 0.3Random crossoverN: 41; end of study: 40Baseline: 68.2 kg (10)
Home, 198288 Month 2.5: 68.9 (9.2) 67 (9.2), p = 0.023 –1.9Random crossoverN: 11, end of study: 10Baseline: 67.6 kg (8.5)
Schmitz, 198986 Month 6: 73 (17.8) 72.8 (18.4), p = ns –0.2Random crossoverN: 10Baseline: 71.6 kg (17.2)
Ziegler, 199084 Final weight not reported Final weight not reportedParallel RCT Average weight gain after Average weight gain afterCSII N: 49; end of study: 36 2 years = 1.3 2 years = 2.8MDI N: 47; end of study: 37 BMI (months 6–24) 22.5 kg/m2 BMI (months 6–24) 24.2 kg/m2,Baseline BMI: 22.1 kg/m2 p < 0.05
At 12 months of follow-up, Bode and colleagues81
found no significant difference in body weightbetween CSII and MDI (68.1 versus 67.4 kg). Thestudy also reported mean weight for CSII only at24 (69.7 kg, SD 15.2), 36 (70.0 kg, SD 14.6) and48 months (68 kg, SD 15.2) of follow-up, but didnot follow MDI for this length of time.
Brinchmann-Hansen and colleagues78 reportedthat weight had increased from baseline by 1.9and 3.4 kg for CSII and MDI, respectively, at24 months of follow-up. No significant differencewas found between CSII and MDI (70.5 versus75.1 kg). Ziegler and colleagues84 also reported a non-significant mean weight gain from baseline (CSII 1.3, MDI 2.8 kg) at 24 months offollow-up. However, body mass index (BMI) wassignificantly lower in CSII than MDI (22.5 versus24.2 kg/m2, p < 0.05) after 6–24 months of follow-up, with a difference in BMI from baselineof –0.5 and +2.1 kg/m2 for CSII and MDI,respectively.84
SummaryThere was no consistent difference in weightbetween CSII and MDI treatment.
Lipid levelsThe majority of studies included in this review didnot report data on cholesterol or triglyceride levels(Table 8). Home and colleagues88 found nosignificant difference in cholesterol after 10 weeksof either CSII or MDI [5.0 mmol/l (SD 0.6) versus5.2 mmol/l (SD 0.6)].
Schiffrin and colleagues91 found cholesterol to besignificantly lower after 6 months of both CSIIand MDI compared with baseline [CSII 163 mg/dl(SD 15) versus 190 mg/dl (SD 9), p < 0.01; MDI
162 mg/dl (SD 12) versus 183 mg/dl (SD 8),p < 0.01]. The difference between CSII and MDIwas not statistically significant. Similarly, nodifference in triglyceride levels was found betweenCSII and MDI [82 mg/dl (SD 14) versus 81 mg/dl(SD 12), p = ns].
Ziegler and colleagues84 did not report data oncholesterol levels but stated that cholesterol levelsincreased within the normal range during thestudy without differences between groups.
SummaryThere is little evidence with which to compare theeffects on cholesterol levels of CSII and MDI.
Patient preferenceStudies reported various forms of patientpreference information, from patients’ treatmentpreference to reasons for preference such ascontrol, lifestyle, appearance or other reason.
Of the eight studies80,82,83,86–88,90,91 reportingpatient preference (Table 9), four studies80,82,83,88
reported a greater proportion of patientspreferring CSII to MDI treatment, rangingbetween 44% and 72.5%. Schiffrin andcolleagues91 reported no difference in the numberof patients preferring to continue with CSII orMDI, with 12.5% of patients reverting toconventional treatment at the end of the study.However, three studies86,87,90 reported a greaterproportion of patients (57%–80%) preferring MDIto CSII. Two further studies84,89 did not reportpatient preference but provided opinions ofpatients regarding treatments.
Reasons for preferring CSII included easierglycaemic control and greater flexibility of
Health Technology Assessment 2004; Vol. 8: No. 43
25
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
TABLE 8 Cholesterol and triglyceride levels in adults with Type 1 diabetes
Study Mean cholesterol and triglyceride level (SD) Difference
CSII MDI(MDI – CSII)
Home, 198288 CholesterolRandom crossover 2.5 months: 5 (0.6) mmol/l 5.3 (0.6) mmol/l 0.3 mmol/lN: 11; end of study: 10
Schiffrin, 198291 CholesterolRandom crossover 6 months: 163 mg/dl (15) 162 mg/dl (12), p = ns –1 mg/dlN: 20; end of study: 16 Triglyceride
6 months: 82 mg/dl (14) 81 mg/dl (12), p = ns –1 mg/dl
Ziegler, 199084 Cholesterol levels increased within the normal range during the study without differences Parallel RCT between treatment groupsCSII N: 49; MDI N: 47
Clinical effectiveness
26
TABLE 9 Patient preference and quality of life for adults with Type 1 diabetes
Study Patient preference/QoL
CSII MDI
Bak, 198787
Random crossover.N: 20; end of study: 16
Preferring: 20%Reasons for preference: freedom frominjections: 44%
75% discomforted by butterfly needle (onCSII), 56% discomforted by infuser itself athome, 21% discomforted by infuser at work
QoL: not reported
80%
95% found pen easy to handle, 30% highfrequency of injections a disadvantage
Of total: independence of fixed mealtimes anadvantage with MDI and CSII by 85%
Chiasson, 198482
Non-random crossoverN: 12
Number preferring: 7 (58%)
QoL: not reported
5 (42%)
Haakens, 199083
Non-random crossoverN: 52; end of study: 35
Number preferring: 23/52 (44%) preferred CSII
Reasons:ControlFlexibility in terms of food intake allowedAvoid frequency injections: 2 (9%)Greater sense of well-being: 2 (9%)
QoL: not reported
20/52 (38%) preferred MDI
Reasons:Control: 17 (85%)More flexible lifestyle than CT without needto wear a devicePreferred human isophane at bedtime:16 (46%)Preferred human ultralente at bedtime:7 (20%)No preference between bedtime insulin:11 (31%)
Hanaire-Broutin, 200080
Random crossoverN: 41; end of study: 40
Number preferring: 29 (72.5%)Of these 29, 21 previously on CSII + regularinsulin and 8 on MDI
QoL: not reported
11 (27.5%)Of these 11, 10 previously on CSII and 1MDI
Home, 198288
Random crossoverN: 11; end of study: 10
8 (80%) preferred pump to MDI, 4 wished tocontinue with it. Universally regarded as toolarge and uncomfortable
QoL: not reported
None wished to continue with 3 dailyinjections owing to inconvenience
Nathan, 198289
Random crossoverN: 5
Number preferring: not reported
Ease of administering and adjusting insulin dosesnoted. A decrease in hypos and improvedfeeling of well-being contributed to acceptanceof pump.
Drawbacks: difficulty for women in dressing toaccommodate pump, and removal of pump forwatersports
QoL: not reported
Saurbrey, 198890
Random crossoverN: 21; end of study: 19
Number preferring for future treatment: 6(31.5%)
Reasons: freedom in daily life: 11 (42%)
QoL: not reported
12 (63%)1 (5%) was unsure of preference (either ICTor CSII)
Reasons: freedom in daily life: 11 (58%)
continued
meals83,91 and freedom from injections.87 Haakensand colleagues83 reported that two of the23 patients preferring CSII thought it wasimportant to avoid frequent injections and two felt a sense of greater well-being, whereas thepatients who preferred MDI found it practical andeasy to administer and that it allowed a moreflexible lifestyle than conventional treatmentwithout the need to wear a pump. Nathan andcolleagues89 did not report levels of patientpreference, but noted ease of administration and adjustment of insulin with CSII, and stated that a decreased number of hypoglycaemicepisodes and improved feeling of well-beingcontributed to acceptability of the CSII. Saurbreyand colleagues90 reported that 58% of patients felt that MDI gave the greatest freedom in dailylife, compared with 42% who felt that CSII gavemore freedom. Bak and colleagues87 reported that 85% of patients found the independence ofmealtimes with both CSII and MDI to beadvantageous.
Home and colleagues88 reported that eight of 10 patients preferred CSII to MDI and four ofthese wanted to continue with CSII rather thanreturn to conventional treatment. However theauthors noted that the pump was generallyregarded as too large and uncomfortable.Similarly, the reasons given by patients in thestudy by Schiffrin and colleagues for notpreferring CSII include the size of the device.However, both of these studies were published in 1982 when pumps were much larger than the present devices. Other drawbacks with thepump included discomfort from the needle orinfuser,87 problems of dressing to accommodatethe pump for women and in watersportsparticipation89 and discomfort, distortion of body image and feeling of dependency on thedevice.91
Ziegler and colleagues84 do not report level ofpatient preference, but do report that fourpatients changed to MDI from the CSII arm
Health Technology Assessment 2004; Vol. 8: No. 43
27
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
TABLE 9 Patient preference and quality of life for adults with Type 1 diabetes (cont’d)
Study Patient preference/QoL
CSII MDI
Schiffrin, 198291
Random crossoverN: 20; end of study: 16
Number preferring: 7 (44%)
7/16 (44%) of pts chose CSII for easierglycaemic control and also greater flexibility formeal hours and meal size
CSII rejected for reasons of bulk, discomfort, afeeling of dependency on the machine anddistortion of body image
70% would recommend CSII to a diabeticfriend
QoL: not reported
7 (44%)[2 pts reverted to conventional]
30% would recommend MDI to a diabeticfriend
Schmitz, 198986
Random crossoverN: 10
Number preferring: 4 (40%)
QoL: not reported
6 (60%)
Tsui, 200179
Parallel RCTCSII N:13; end of study: 12MDI N: 14 end of study: 14
Number preferring: not reported
QoL score: (mean)Satisfaction 75.6Impact: 69.9Diabetic worry: 85.2Social worry: 89.6Global health: 68.2Response rate 85%
QoL score (mean):68.368.479.894.067.3 (p = ns for any of the subscales)Response rate 100%
Ziegler, 199084
Parallel RCTCSII N: 49; end of study: 36MDI N: 47; end of study: 37
pts, Patients.
Number preferring: not reported
4 pts changed to MDI during study owing totechnical problems with pump
QoL: not reported
7 pts changed treatment group during studyowing to explicit request to have a pump
owing to technical problems with the pump, andseven MDI patients explicitly requested a pumpduring the study.
SummaryFour of the eight studies reported that a greaterproportion of patients preferred CSII, threestudies found a greater proportion preferred MDIand one study found no difference. However, sincesome of these studies were done with older andbulkier pumps, their findings may not be relevantto today’s pumps. Progress has also been madewith MDI, for example the introduction of discreet‘pen’ injectors. The three studies that reportedthat greater proportions preferred MDI were from 1987 to 1989; those where the majoritypreferred CSII were from 1982, 1984, 1990 and 2001.
Quality of lifeOnly one study (Tsui 2001)79 reported on QoL(Table 9). Information was collected using theDiabetes Quality of Life (DQOL) tool. Twoparticipants did not complete the DQOL, givingresponse rates of 85% for CSII and 100% for MDI,and just three of 1128 items were missing. Nosignificant differences between CSII and MDI werefound overall or on any of the subscales:satisfaction (7.3), impact (1.5), diabetic worry (5.4),global health (0.9) or social worry (–4.4).
SummaryThere is little evidence with which to compare theeffects of CSII and MDI on QoL.
Adverse effectsThe definition of hypoglycaemic episodes varied,making comparison between studies difficult.Significantly fewer hypoglycaemic episodes with CSII therapy were observed in threestudies,81,85,87 but most studies found no statisticaldifference between CSII and MDI78–80,83,89–91
(Table 10). Ziegler and colleagues reportedsignificantly more episodes of blood glucose<50 mg/dl and reported symptoms ofhypoglycaemia with CSII than MDI at 1–6 and7–12 months of follow-up, but these tended todecrease in CSII so that at 19–24 months offollow-up the differences were no longersignificant.84
Severe hypoglycaemic episodes differed littlebetween treatments in most studies,79,80,83,85–88,90,91
with no severe episodes occurring in twostudies.86,90 Significantly fewer severehypoglycaemic episodes per 100 patients years occurred with CSII compared with
MDI in the study by Bode and colleagues,81
and similarly Brinchmann-Hansen and colleagues found hypoglycaemic coma occurredless frequently with CSII.78 Ziegler and colleagues, however, found that hypoglycaemiccoma was about twice as common in CSII than MDI.84
Two studies reported no occurrences of DKA with either treatment,86,87 and another threestudies reported no events with MDI78,83,90
and one83,90 or two78 episodes with CSII. Bodeand colleagues found rates of DKA to be less with CSII than MDI, although the differencewas not significant (7.2 versus 14.6 events per 100 patient years). Nosadini and colleagues,however, reported significantly more ketotic eventswith both fixed-rate and variable rate CSIIcompared with MDI. About 20% of these episodeswere caused by pump malfunction, and 23% ofthose with fixed basal rate and 27% of those withvariable basal rate were caused by infectivedisease.85 Ziegler and colleagues found anincidence of DKA of 9.7 events per 100 years inpatients assigned to CSII and 8.1 events per100 years in patients assigned to MDI; however,when the true therapy at the time of the event was analysed, more events occurred with CSII(14.1 versus 2.9 events per 100 years). However, it should be noted that the studies thatreport significant problems with DKA are theolder ones.
The presence or absence of other adverse events was poorly reported in many of the studies.Pump malfunctions were reported in three studies, specified as the cause of about 20% ofketotic episodes by one study,85 but occurring only once88 or twice89 in the other studies.Problems with catheters were reported in onestudy, in which the catheter became dislodged atnight on six occasions.89 Subcutaneous abscessoccurred in one patient with CSII in twostudies83,89 and abscesses occurred in eightpatients with CSII in the study by Brinchmann-Hansen and colleagues.78 Subcutaneous infectionoccurred three times in one patient in anotherstudy.90
CaveatsIn most of the RCTs, there was little difference in the frequency of severe hypoglycaemia. Thisevidence is in conflict with that from three othersources, namely observational studies, clinicalopinion and patient experiences reported to usduring this review. The observational studiesinclude:
Clinical effectiveness
28
Health Technology Assessment 2004; Vol. 8: No. 43
29
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
TABLE 10 Adverse events for adults with Type 1 diabetes
Study Adverse events
CSII MDI
Bak, 198787
Random crossoverN: 20; end of study: 16
Frequency of BG values <4 mmol/l: 2.3%Severe hypos: 1 (requiring hospital treatment)DKA: 0
Frequency of background values <4 mmol/l:4.1%, p < 0.02
Severe hypos: 0
DKA: 0
Bode, 199681
Non-random crossoverN: 55
Severe hypos: 12 episodes (12 months)
Severe hypos per 100 patient-years (n),significance from MDI: 12 months: 22 (55) p < 0.000124 months: 26 (50) p < 0.0136 months: 39 (33) p < 0.000148 months: 36 (25) p < 0.01
Of 25 pts with HbA1c ≥ 8.0%, rate of severehypos declined from 84 events per 100 patientyears at baseline (MDI) to 8 events in year 1(p < 0.0001). The 30 pts with baseline HbA1c<8.0% also experienced a decline, from 183 to33 severe hypos per 100 patient years(p = 0.0005).
DKA rates were not significantly differentbetween MDI and CSII phases (14.6 vs 7.2events per 100 patient years, respectively)
Severe hypos: 76 episodes (12 months)Severe hypos per 100 patient years (n):138(55)
Brinchmann-Hansen,198878
Parallel RCT.CSII N: 15; MDI N: 15
Hypoglycaemia: mean (SEM):% of home BG <2.5 mmol/l: at randomisation= 6 (2); after 2 years = 11 (2)
Symptomatic episodes/week/patient: At randomisation = 2.3 (0.6); at 2 years = 1.7(0.3)
Hypo coma: 2 (2)DKA: 2 (2)Subcutaneous abscess: 8 (6)
Retinopathy: improved (14%), unchanged(29%), worsened (57%) (one patient had anallergy to fluorescein)
Hypoglycaemia: mean (SEM):% of home BG <2.5 mmol/l: atrandomisation = 8 (3); after 2 years = 7 (1),vs CSII p = ns
Symptomatic episodes/week/patient: At randomisation = 1.5 (0.3); at 2 years =1.2 (0.2), vs CSII, p = ns.Hypo coma 14 (6), vs CSII p < 0.001.DKA 0, vs CSII p = nsSubcutaneous abscess 0, vs CSII p < 0.01
Retinopathy: Improved (7%), unchanged(43%), worsened (50%) (one patient had anallergy to fluorescein)
Haakens, 199083
Non-random crossoverN: 52; end of study: 35
Subcutaneous abscess: 1
DKA: 1 requiring hospitalisationSevere hypos (episodes/patient/month):Total 1/7.5 (requiring glucose i.v./glucagon 1/96;assistance but not requiring glucose i.v./glucagon1/8).
No. of mild subjective hypos/week: 1.4 (SD0.92)
Subcutaneous abscess: 0
DKA: 0
Severe hypos (episodes/patient/months): Total 1/5 (requiring glucose i.v./glucagon 1/79;assistance but not requiring glucosei.v./glucagon 1/5). CSII vs MDI p = ns.
No. of mild subjective hypos/week: isophane1.4 (SD 0.92), ultralente 1.5 (SD 1.17). CSIIvs MDI p = ns.
Hanaire-Broutin, 200080
Random crossoverN: 41; end of study: 40
Severe hypos: 2 pts, 3 events (severe hypos didnot result in coma or seizures, external helpneeded to take sugar, but glucagon or glucoseinjection not required)
Hypos during last 14 days of each treatmentperiod: 3.9 (4.2)
Severe hypo: 1 patient, 1 event
Hypos during last 14 days of each treatmentperiod: 4.3 (3.9), p = ns
continued
Clinical effectiveness
30
TABLE 10 Adverse events for adults with Type 1 diabetes (cont’d)
Study Adverse events
CSII MDI
Home, 198288
Random crossover.N: 11; end of study: 10Baseline hypo 1.1 (0–4)episodes per pt biweekly
Hypos: 0.6 (0–3) episodes per pt biweekly.Pump malfunction: 1 (1 pt × 3 hypo episodes in16 h due to faulty pump)
1 patient required intravenous glucose
Hypos: 0.8 (0–7) episodes per pt biweekly
Nathan, 198289
Random crossoverN: 5
During 2.5 patient years CSII: subcutaneousabscess needing oral antibiotics (1), pumpmalfunctions (2) catheter dislodged at night (6)
Hypo reactions n/week at 3 months: 1 (0.25)
Hypo reactions n/week at 3 months: 2.5(0.61), p = ns
Nosadini, 198885
Parallel RCTCSII fixed N: 19;MDI N: 15;CSII variable N: 10
Mean events/pt/yr:CSII with fixed night basal rate:Hyper: 18 (5) vs MDI, p = nsKetotic events: 0.13 (0.02) vs MDI, p < 0.01Mild hypos: 36 (10) vs MDI, p < 0.01Severe hypos: 0.14 (0.05) vs MDI, p < 0.01Sudden death: 0.04 (1 pt dead in bed)
CSII with variable night basal rate):Hyper: 17 (4) vs MDI, p = nsKetotic events: 0.16 (0.03) vs MDI, p < 0.05Mild hypos: 30 (11) vs MDI, p < 0.01Severe hypos: 0.16 (0.09) vs MDI, p < 0.01[CSII–HOR vs CSII–FBR, all p = ns]
Malfunction of pump caused about 20% ofketotic episodes in both CSII groups. 23% (CSIIfixed) and 27% (CSII variable) caused byinfective disease
Hyper: 20 (3) Ketotic events: 0.03 (0.01) Mild hypos: 59 (12) Severe hypos: 0.42 (0.15)
Saurbrey, 198890
Random crossoverN: 21; end of study: 19
Severe hypos: 0
No. of subjective or biochemical (BG<2.5 mmol/l) hypos in each group, p = ns(values not given).
Ketoacidosis = 1 (due to acute gastroenteritisand failure to take extra insulin, not pumpmalfunction)
Subcutaneous infection (3 times) in samepatient, one required surgical incision
Severe hypos: 0
Schiffrin, 198291
Random crossoverN: 20; end of study: 16
Mild hypos: 186Moderate hypos: 10Severe hypos: 1 (due to delaying meal afterbolus insulin taken – intravenous glucose wasadministered)
Mild hypos: 189, p = nsModerate hypo: 11, p = nsSevere hypos: 0, p = ns
Schmitz, 198986
Random crossoverN: 10
Severe hypos: 0DKA: 0
Severe hypos: 0DKA: 0
Tsui, 200179
Parallel RCTCSII N:13; end of study: 12MDI N: 14; end of study: 14
Severe hypos: 6
Mean number of hypos each month, 8.9(3 months), 7.2 (6 months), 7.0 (9 months), 8.0 (overall)
Severe hypos: 4, p > 0.10
Mean number of hypos each month, 5.0(3 months), 9.0 (6 months), 9.2 (9 months),7.4 (overall), all = ns vs CSII
Relative treatment effect (CSII-MDI)/MDI foroverall no. of hypos: +9% (95% CI –37 to+87), p > 0.10
continued
� Boland and colleagues,98 who reported a dropin hypoglycaemic events requiring assistancefrom 134 per 100 person years on MDI to 76 per 100 person years on CSII, inadolescents.
� Mack-Fogg and colleagues,99 in abstract only,report a reduction in severe hypoglycaemicevents in children after transfer from MDI toCSII, from a mean of 0.55 to 0.25 per child peryear.
� Rudolph and Hirsch100 found that there were73 severe hypoglycaemic events per 100 patientyears on MDI, but only 19 on CSII.
� Haardt and colleagues,101 in a retrospectiveanalysis of the Hotel–Dieu cohort, found that glycated haemoglobin levels decreased‘only modestly’ (from 9.3% on injections to 8.6% on CSII), but that the number of severe hypoglycaemic events fell by over 75%.
A possible explanation is that patients recruited tothe trials may be roughly representative of thoseseen in clinics, although with a bias to the mostcooperative (because of the demands of trials) whomay have better control than average, whereas inroutine care as reflected more in the observationalstudies, CSII may be used mostly in those who arehaving considerable problems with control, eitherhigh glycated haemoglobin or frequenthypoglycaemic events.
SummaryHypoglycaemic events did not differ significantlybetween CSII and MDI in most trials, but threefound fewer hypoglycaemic episodes with CSIIand one study found more hypoglycaemia andhypoglycaemic coma with CSII. Observationalstudies found much greater reductions in thefrequency of severe hypoglycaemia with CSII, butare more prone to bias.
PregnancyFour parallel RCTs92–95 compared the effects ofCSII with MDI in pregnancy (Tables 4 and 11–14,Appendices 7 and 11), although the sample size inthree of the studies was small (14–32).92–94 Cartaand colleagues94 included women with Type 1 andType 2 diabetes diagnosed before pregnancy andanalysed these groups separately. Nosari andcolleagues92 and Coustan and colleagues93
included women with Type 1 diabetes who werepregnant or planning pregnancy, and Burkart andcolleagues included women with Type 1 diabetesattending no later than the first trimester.95
Glycated haemoglobinBurkart and colleagues did not report data onglycated haemoglobin, simply stating that patientsin both groups had normal mean glucose andHbA1 <7.5% at least from the end of the firsttrimester on.95 During the first trimester in theother studies, glycated haemoglobin was lower
Health Technology Assessment 2004; Vol. 8: No. 43
31
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
TABLE 10 Adverse events for adults with Type 1 diabetes (cont’d)
Study Adverse events
CSII MDI
Ziegler, 199084
Parallel RCTCSII N: 49; end of study:36.MDI N: 47; end of study: 37
FBR, fixed basal rate; HOR, higher overnight rate; SEM, standard error of mean.
No. of metabolic complications/6 months mean(range). Reviewer has calculated mean of the 4 × 6 month periods to give totals below.Mild hypos (BG <50 mg/dl): 15.5 (0–70)Mild hypos (symptoms): 19.8 (0–104)Severe hypos (req. assistance): 0.091 (0–7)Severe hypos (coma): 0.101 (0–4)Ketosis (BG >200 mg/dl and ketonuria): 6.5(0–57)Ketoacidosis: 0.046 (0–1)
Frequency of pts (no. of events/100 years):Hypos requiring assistance: 13.9 (18.1)Hypos (coma): 30.6 (19.4)DKA: 13.9 (9.7) [true therapy at time of event:15.4 (14.1)]
Mild hypos (BG <50 mg/dl): 10.6 (0–53)Mild hypos (symptoms): 10.1 (0–81)Severe hypos (requiring assistance): 0.070(0–2)Severe hypos (coma): 0.048 (0–1)Ketosis (BG >200 mg/dl and ketonuria): 7.8(0–85)Ketoacidosis: 0.040 (0–3)
Frequency of pts (no. of events/100 years):Hypos requiring assistance: 16.2 (13.5)Hypos (coma): 13.5 (9.5)DKA: 8.1 (8.1) [true therapy at time ofevent: 5.9 (2.9)]
with CSII than MDI (0.2–1.1%) in both Type 1diabetes92,94 and Type 2 diabetes,94 but thesedifferences were not statistically significant (Table11). Similarly, there were no significant differencesbetween CSII and MDI during the second trimester(–0.7 to 0.75%)92,94 or term (–0.1 to 0.7%).94
Insulin doseThere was no significant difference between CSIIand MDI in total daily insulin dose in Type 1diabetes92,94 or Type 2 diabetes94 during the first,second or third trimesters (Table 12). However,Coustan and colleagues93 found that fewer insulinunits per kilogram per day were required withCSII than with MDI [first trimester, 0.71 U/kg/day(SD 0.16) versus 1.01 U/kg/day (SD 0.28), p = 0.101; second trimester 1.02 U/kg/day (SD0.53) versus 1.40 U/kg/day (SD 0.4), p = 0.027;third trimester, 1.26 U/kg/day (SD 0.49) versus1.63 U/kg/day (SD 0.51), p = 0.041]. Daily insulin
dose was not reported by Burkart andcolleagues.95
Pregnancy outcomesThe mean duration of pregnancy was38.1–38.9 weeks in CSII and 37.7–38.8 weeks inMDI, with no significant differences between thegroups for Type 1 or Type 2 diabetes92–95
(Table 13). Nosari and colleagues92 reported threeintrauterine deaths: two in the CSII group andone in the MDI group. In one case from eachgroup, the cause of death was a severe reductionin placental blood flow; the cause of the thirddeath was unknown. Burkart and colleaguesreported two deaths in the CSII group, oneintrauterine death due to cause unknown and onedeath at 3 months postpartum followingpremature delivery and maternal ketoacidosis.Three deaths occurred in the MDI group at3 days, 5 days and 3 months postpartum due to
Clinical effectiveness
32
TABLE 11 Glycated haemoglobin in pregnant women
Study Mean glycated haemoglobin (%) (SD) Difference
CSII MDI(MDI – CSII)
Burkart, 198895 Data not reported. States that patients selected in the CSII or MDI group had normal Parallel RCT mean glucose and HbA1 <7.5% at least from the end of the first trimester on.CSII N: 48; MDI N: 41Baseline not reported
Carta, 198694 Weeks 9–11: 7.8 8.5, p = ns 0.7Parallel RCT Weeks 19–21: 7 7.75, p = ns 0.75Type 1 Weeks 33–35: 7 6.95, p = ns –0.05CSII N: 8; MDI N: 7 Weeks ≥ 37: 6.5 6.45, p = ns –0.05HbA1 data collectedBaseline CSII: 8.75%Baseline MDI: 9.1%, p = ns
Carta, 198694 Weeks 9–11: 7 8.1, p = ns 1.1Parallel RCT. Weeks 19–21: 6.75 7.2, p = ns 0.45Type 2 Weeks 33–35: 6.5 6.7, p = ns 0.2CSII N: 6, MDI N: 8. Weeks ≥ 37: 6 6.7, p = ns 0.7HbA1 data collectedBaseline CSII: 8.1%Baseline MDI: 8.2%, p = ns
Coustan, 198693 Therapy week 8: 7 7.5, p = ns 0.5Parallel RCT Term: 6.3 (0.6) 6.4 (0.4), p = ns 0.1CSII N: 11; MDI N: 11HbA1 data collectedBaseline CSII: 8.6% (1.7)Baseline MDI: 9.1% (1.5), p = ns
Nosari, 199392 1st trimester: 6 (3.6) 6.2 (1.6), p > 0.5 0.2Parallel RCT 2nd trimester: 6.8 (5.6) 6.1 (2.4), p > 0.5 –0.7CSII N: 16; MDI N: 16 3rd trimester: 6.3 (2) 6.2 (0.8), p > 0.5 –0.1HbA1c data collectedBaseline not reported
toxoplasma infection, hypoplastic left heart andrenal vein thrombosis, respectively.95 Nointrauterine deaths occurred in the other twostudies.93,94 The proportion of babies delivered byCaesarean section varied between the studies, buttended to be similar between CSII and MDI inType 1 diabetes. Carta and colleagues,94 however,reported that in Type 2 diabetes, one of nine(16.7%) women with CSII underwent Caesareansection compared with three of eight (37.5%)women with MDI. There was no significantdifference in birthweight between CSII and MDI in any of the studies. A small number ofpreterm, small for gestational age or large forgestational age infants were observed in both groups, with little apparent differencebetween the groups.
Patient preference and quality of lifePatient preference and QoL were not reported bythe four studies comparing CSII and MDI inpregnancy.
Adverse eventsNo pump malfunctions were reported, althoughone study reported three catheter disconnections94
and another stated that catheter leakage or
occlusion occurred infrequently and resolvedquickly93 (Table 14). Carta and colleagues94
reported that among Type 2 diabetes patients,episodes of mild hypoglycaemia were few and didnot occur in all patients. One severehypoglycaemic episode occurred among the Type1 diabetes patients, but the treatment group wasnot specified. Coustan and colleagues93 foundmore severe hypoglycaemic episodes in the MDIgroup than in the CSII group (45% versus 27%),although the authors noted that the groups weretoo small to attain statistical significance. Twoother studies, however, found slightly more severeepisodes92 and undefined hypoglycaemia95 withCSII. Ketoacidosis occurring in CSII group butnot the MDI group was reported by two studies.92
It is important to bear in mind that even oneepisode of ketoacidosis could lead to fetal death.95
SummaryThere is insufficient evidence from these studiesthat CSII is better than MDI in pregnancy forachieving glycaemic control. If CSII or any otherform of intensified therapy is to be optimallyeffective in pregnancy, it would have to startbefore conception, partly to minimise risk of
Health Technology Assessment 2004; Vol. 8: No. 43
33
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
TABLE 12 Insulin dose in pregnant women
Study Mean insulin dose (SD) Difference
CSII MDI(MDI – CSII)
Carta, 198694 Month 3: 37 U/day 32 U/day, p = ns –5 U/dayParallel RCT Month 6: 42 U/day 47 U/day, p = ns 5 U/dayType 1 Month 9: 49.1 U/day 60.4 U/day, p = ns 11.3 U/dayCSII N: 8; MDI N: 7Baseline CSII: 31.5 U/dayBaseline MDI: 27.8 U/day, p = ns
Carta, 198694 Month 3: 23.1 U/day 28 U/day, p = ns 4.9 U/dayParallel RCT Month 6: 36 U/day 49 U/day, p = ns 13 U/dayType 2 Month 9: 52.8 U/day 51.2 U/day, p = ns –1.6 U/dayCSII N: 6; MDI N: 8Baseline not reported
Coustan, 198693 1st trimester: 0.71 U/kg/day (0.16) 1.01 U/kg/day (0.28), p = 0.101 0.3 U/kg/dayParallel RCT 2nd trimester: 1.02 U/kg/day (0.53) 1.4 U/kg/day (0.4), p = 0.027 0.38 U/kg/dayCSII N: 11; MDI N: 11 3rd trimester: 1.26 U/kg/day (0.49) 1.63 U/kg/day (0.51), p = 0.041 0.37 U/kg/dayBaseline CSII: 0.9 U/kg/day (0.42)Baseline MDI: 1.05 U/kg/day (0.38), p = 0.381
Nosari, 199392 1st trimester: 39.6 U/day (22.8) 36 U/day (24.4), p > 0.5 –3.6 U/dayParallel RCT 2nd trimester: 48.1 U/day (20.8) 43.7 U/day (35.6), p > 0.5 –4.4 U/dayCSII N: 16; MDI N: 16 3rd trimester: 57.3 U/day (21.6) 54.7 U/day (39.2), p > 0.5 –2.6 U/dayBaseline not reported
Clinical effectiveness
34
TABLE 13 Outcomes of pregnancy
Study CSII MDI
Burkart, 198895
Parallel RCT
CSII N: 48MDI N: 41
Duration: mean 38.4 weeks Live births: 47/48 (97.9%)Birth weight: mean 3082.7 (units not specified)No. Caesarean: 24/48 (50%)
Complications in pregnancy:Pyelonephritis 5Premature labour 4 Premature rupture 2Premature delivery 7 Pre-eclampsia 3Growth retardation 3Hypoglycaemia 5 Still birth 1 Ketoacidosis 1Total: 12/48 pregnancies had one or morecomplications
Fetal outcome:Healthy newborns 40Minor symptoms 8 Major symptoms 0Glucose i.v. 14Overall mortality up to age 1 year = 2/48 (4.2%)
Duration: mean 38.2 weeksLive births: 100%Birth weight: mean 3319.5 (units not specified)No. Caesarean: 15/41 (36.6%), p = ns
Complications in pregnancy:Pyelonephritis 6Premature labour 3Premature rupture 1Premature delivery 4Pre-eclampsia 3 Growth retardation 1Hypoglycaemia 3Still birth 0Ketoacidosis 0Total: 13/41 pregnancies had one or morecomplications (vs CSII p = ns)
Fetal outcome:Healthy newborns 34Minor symptoms 5Major symptoms 2 Glucose i.v. 9 (all p = ns vs CSII)Overall mortality up to age 1 year = 3/41(7.3%), vs CSII p < 0.05a
Incidence of complications in both groups islinked to the severity of maternal diabetes(White’s criteria). Birthweight reportedaccording to White’s criteria decreases withseverity. Differences in mean birthweights, p = ns.
Carta, 198694
Parallel RCT
Type 1CSII N: 8MDI N: 7
Duration: mean 266.7 days (SD 7.6)Live births: 100%Birth weight: mean 3395 g (SD 407)No. Caesarean: 3 (37.5%)
1 preterm due to hydramnios and fetalmegalosomnia. 1 large for gestational age. 1 macrosomnic (>4000 g).[CSII or MDI group not specified: 2 infantsdepressed, 5 metabolic morbidity (hypoglycaemia,hypocalcaemia, hyperbilirubinaemia), 2 congenitalcardiac malformations]
Duration: mean 263.6 days (SD 17.7)Live births: 100%Birth weight: mean 2906 g (SD 553), CSII vsMDI p = nsNo. Caesarean: 3 (42.9%)
1 preterm due to metrotthagia caused by centralplacenta previa
Carta, 198694
Parallel RCT
Type 2CSII N: 6MDI N: 8
Duration: mean 272 days (SD9)Live births: 100%Birth weight: mean 3292 g (SD 578)No. Caesarean: 1 (16.7%)
1 preterm due to premature rupture of membranes.1 large for gestational age
Duration: mean 269 days (11.7)Live births: 100%Birth weight: mean 2994 g (SD 512), CSII vsMDI p = nsNo. Caesarean: 3 (37.5%)
1 preterm due to premature rupture ofmembranes. 1 small for gestational age
Coustan, 198693
Parallel RCT
CSII N: 11MDI N: 11
Duration: mean 38.1 weeks (SD 2.1)Live births: 100%Birth weight: mean 3050 g (SD 675)No. Caesarean: 7 (63.6%)
One birthweight above 90th percentile, 2 small forgestational age babies. Neonatal hypoglycaemia inone
Duration: mean 38.8 weeks (SD 1.4)Live births: 100%Birth weight: mean 3324 g (SD 475), CSII vsMDI p = nsNo. Caesarean: 7 (63.6%)
1 baby moderate respiratory distress syndrome.Neonatal hypoglycaemia in one
continued
Health Technology Assessment 2004; Vol. 8: No. 43
35
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
TABLE 13 Outcomes of pregnancy (cont’d)
Study CSII MDI
Nosari, 99392
Parallel RCT
CSII N: 16MDI N: 16
a The p-value reported in the paper appears to be incorrect, and should be statistically insignificant.
Duration: mean 38.9 weeks (SD 7.6)Live births: 87.5%Birth weight: mean 3130 g (SD 1480)No. Caesarean: 9 (56.3%)
2 intrauterine deaths. No fetal congenitalmalformations, macrosomia or low Apgar scores(<7 at 5 minutes)1 CSII large for age. Respiratory distress in 1 CSIIinfant. Neonatal hypo (plasma glucose <30 mg/dl) inone CSII
Duration: mean 38.2 weeks (SD 10), CSII vsMDI p > 0.5Live births: 93.8%Birth weight: mean 3010 g (SD 1728), CSII vsMDI p > 0.5No. Caesarean: 7 (43.8%)
1 intrauterine death. 2 MDI small for age. 1premature MDI birth. Neonatal hypo (plasmaglucose < 30 mg/dl) in one MDI
TABLE 14 Adverse events in pregnant women
Study details CSII MDI
Burkart, 198895
Parallel RCT
CSII N: 48MDI N: 41
Hypoglycaemia: 5Ketoacidosis: 1
Hypoglycaemia: 3 Ketoacidosis: 0
Coustan, 198693
Parallel RCT
CSII N: 11MDI N: 11
Injection site infection: 0Pump malfunction: 0Mild hypos: 11 patientsModerate hypos: 6 patientsSevere hypos: 3 patients (8 episodes)DKA: 0Other: 1 hyperglycaemia and ketonuria (2 days)during viral infectionCatheter leakage or occlusion occurred infrequentlyand resolved quickly
Mild hypo: 11 patientsModerate hypos: 10 patientsSevere hypos: 5 patients (23 episodes)DKA: 0
one subject had 17 hypos related to dietaryirregularities
Nosari, 199392
Parallel RCT
CSII N: 16MDI N:16
Severe hypos: 3DKA: 3
Severe hypos: 1, CSII vs MDI p > 0.5DKA: 0, CSII vs MDI p > 0.1
Carta, 198694
Parallel RCT
Type 2CSII N: 6MDI N: 8
Episodes of mild hypos were few and did not occur in all patients
Carta, 198694
Parallel RCT
Type 1CSII N: 8MDI N: 7
Injection site infection: 0Pump malfunction: 0Other: catheter disconnection: 3In all pts, no hypo coma. One severe hypo (group not stated). Mild hypos occurred on average onceevery 15–20 days.
malformation and partly because it may take a fewmonths to learn how to use the intensifiedregimen.
AdolescentsTwo randomised crossover studies compared theeffects of MDI and CSII in adolescents aged up to20 years with Type 1 diabetes,96,97 one of whichalso included treatment with CSII overnight plusMDI during the day (CSII + MDI)97 (Tables 4, 15and 16, Appendices 8 and 12).
Glycated haemoglobinAfter 1 month, glycated haemoglobin wassignificantly lower than baseline [13% (SD 1),p < 0.001] in all three groups (Table 15) in thestudy by Schiffrin and colleagues.97 At 4 months offollow-up, glycated haemoglobin was significantlylower with CSII (8.8%) than with MDI (9.6%,p < 0.05) or CSII + MDI (9.3%, p < 0.05). Thedifference between CSII + MDI and MDI was notstatistically significant.97 However, althoughTamborlane and colleagues also found a decreasein glycated haemoglobin with both CSII and MDIcompared with baseline (p < 0.05), they found nosignificant difference between CSII and MDI after6 months of follow-up (8.5% versus 8.7%,p = ns).96
InsulinTotal daily insulin dose was significantly lowerduring both CSII (44 U/day) and CSII + MDI(48 U/day) than at baseline [64 U/day (SD 14),p < 0.001] or during MDI (60 U/day, p < 0.001)in the study by Schiffrin and colleagues97
(Table 15). Conversely, Tamborlane and colleagues
found an increase in insulin dose with both CSIIand MDI compared with baseline (p < 0.05), butno significant difference between CSII and MDIafter 6 months of follow-up (1.3 versus 1.4 U/day,p = ns).96
Patient preference and quality of lifeQoL was not reported by either study.96,97 At theend of the study by Schiffrin and colleagues, overhalf of patients (55%) chose to be treated withCSII and all patients stated that they wouldrecommend this treatment to a friend (Table 16).Four patients (20%) chose to continue with MDI and three (15%) with CSII + MDI, and two(10%) patients returned to using twice dailyinjections. MDI and CSII + MDI would berecommended to others by 66% and 70% ofpatients respectively.97
Adverse eventsBoth studies reported few severe hypoglycaemicepisodes96,97 (Table 16). Schiffrin and colleagues97 reported that there was no differencein the number of hypoglycaemic events between treatments, although numbers are not stated, and that they were mostly mild and averaged one per week per patient. Similarly, Tamborlane and colleagues reportedthat episodes of mild hypoglycaemia werecommon during each treatment.96 No otheradverse events were reported by either study.
SummaryOne study found no significant difference betweenCSII and MDI, whereas the other study showed a
Clinical effectiveness
36
TABLE 15 Glycated haemoglobin and daily insulin dose in adolescents
Study Mean HbA1 (SD) (%) Mean daily insulin dose (SD) (U/day)
CSII MDI CSII + MDI CSII MDI CSII + MDI
Schiffrin, 198497 Month 3: 8.75 9.5 9Random crossover Month 4: 8.8 9.6 9.3 Month 4: 44 60 48N: 20Baseline HbA1: 13% (1) CSII vs MDI p < 0.05 CSII vs MDI p <0 .001Baseline daily insulin dose: CSII vs CSII + MDI p < 0.05 CSII + MDI vs MDI p < 0.00164 U/day (14)
Tamborlane, 198996 Month 6: 8.5 (1.26) 8.7 (1.58), p = ns Month 6: 1.3 (3.16) 1.4 (0.32), p = nsRandom crossoverN: 10Baseline HbA1: 12.2% (4.11)Baseline daily insulin dose: 1.1 U/day (0.32)
clear benefit of CSII on glycated haemoglobin andinsulin dose.
CSII in childrenNo published randomised trials were found eitherby our searches or in published reviews.65,102–104
Table 17 summarises results from some case series.Some include both adolescents and children.
Two recent abstracts from the ADA 2002conference were found. The first, by Fox andcolleagues,109 gives only sparse data, and it is notclear what treatment patients were on beforestarting CSII. They may have been newlydiagnosed. Numbers are small, 10 patients todate. This study was excluded. The secondabstract, by Weintrob and colleagues,110 is furtheradvanced, with 23 children aged 9–14 years. It is acrossover study with 3.5 months on each of CSIIand MDI. There was no difference in HbA1c butthere were fewer hypoglycaemic episodes on CSII.
These studies suggest that CSII has a place intreatment of children with diabetes, but case seriesare more susceptible to bias than RCTs and betterevidence is needed. In most cases, children wereswitched directly from CT to CSII. Nowadays,children are more likely to go from MDI to CSII.Some may have problems running out of skin siteswith MDI. One small study in 10 children aged7–10 years studied a combination of CSII fromdinner until dawn and prebreakfast lispro andNPH; the comparator was three-injection MDI.66
The study was very short, with only 4 weeks stable
on each treatment. HbA1c results were notavailable given those durations. Blood glucoseresults improved. For children unable to use 24-hour CSII, this may be a useful option, butfurther research is needed.
Overnight-only CSII in adultsThere is little evidence on overnight-only CSII inadults. Olsson and colleagues in 198768 compareda bedtime injection of intermediate acting insulinwith overnight CSII in 10 patients, but on onlyfour nights. Kanc and colleagues69 carried out a 2-month crossover study and reported lesshypoglycaemia and improved hypoglycaemicepisode awareness.
Short-term CSII in poorly controlledadults with Type 2 diabetesThere is little evidence for this. Valensi andcolleagues111 carried out a short-term study,duration 8–23 days, but CSII was combined withmetformin, and with strict diet, in patientsadmitted to hospital. Little can be deduced aboutCSII from this study.
Another paper from France112 studied all theirType 2 diabetes patients with poor control onmaximal oral therapy. Many needed permanentinsulin treatment, but a group of 111 remained onoral treatment apart from two 3-day periods onCSII, and their HbA1c improved from 8.76% at thestart of the period to 7.82% at the end, with slightweight loss. Hence there may be a place for short-term use of CSII, but further research is required.
Health Technology Assessment 2004; Vol. 8: No. 43
37
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
TABLE 16 Patient preference and adverse events among adolescents
Study CSII MDI CSII + MDI
Patient preferenceSchiffrin, 198497 No. preferring: 11 (55%) No. preferring: 4 (20%) No. preferring: 3 (15%)Random crossover 100% would recommend 66% would recommend MDI to 70% would recommend N: 20 CSII to a friend a friend CSII + MDI to a friend
(2 of 20 completing studyreverted to twice dailyinjections)
Adverse eventsSevere hypos: 1 Severe hypos: 1 (due to Severe hypos: 0
inadvertent administration of extra NPH dose)
Note: No difference in hypos between treatments. Mostly mild and averaged one per week perpatient. Same pt had severe hypos on CSII and MDI.
Tamborlane, 198996 Mild hypos were common during each treatment. States that there was only one severe Random crossover hypo requiring assistance during each 6-month treatment periodN: 10
Analogue versus soluble insulin in CSIIQuantity and quality of researchSix studies (one parallel RCT113 and fiverandomised crossover studies114–118) comparingrapid-acting analogues with soluble insulin in CSIIwere included in the review and are reported inTables 18–23 and Appendices 9 and 13. Bode andcolleagues113 compared two analogue insulins,lispro and aspart, whereas the other studiesincluded insulin lispro. Randomisation andallocation concealment were adequate in theparallel RCT,113 but the method of randomisationwas not reported in the five crossover studies114–118
(Table 18). The similarity of the groups at baselinewas reported in two studies,113,117 and all studiesreported eligibility criteria. Fivestudies113,114,116–118 reported point estimates andmeasures of variability for the primary outcomemeasure; however, Schmauss and colleagues115
reported the mean and standard error of glycatedhaemoglobin for each treatment at crossover andend of study, but did not combine the two periods.Raskin and colleagues117 were the only group toperform ITT analysis. Loss to follow-up wasadequately reported in two studies,115,118 partially
reported (number but not reasons, or vice versa)in one study113 and inadequately reported(numbers not stated for each group) in twostudies.116,117 Loss to follow-up was not reportedby Renner and colleagues.114
Adults with Type 1 diabetesGlycated haemoglobinGlycated haemoglobin was reported in allincluded studies (Table 19), which were combinedin a meta-analysis (Figure 11). Pooling these data(final values and changes from baseline includedas separate subgroups119) using a random effectsmodel (�2 test for heterogeneity 9.09, df = 5,p = 0.11) showed a significant decrease in glycatedhaemoglobin with the analogue insulin lisprocompared with soluble (–0.26%, 95% CI –0.47 to–0.06). The study by Bode and colleaguesincluded a third group with insulin aspart.113
Replacing lispro with aspart for this study in themeta-analysis made little difference to the overalltreatment effect (–0.27%, 95% CI 0.44 to –0.10).
Insulin doseInsulin requirement varied between the studies(Table 20); however, none of the five studiesreporting insulin dose found a significant
Clinical effectiveness
38
TABLE 17 CSII in children: case series
Study No. of Previous Reason for starting HbA1(c) HbA1(c) Discontinuation patients treatment CSII before after (%) rate (%)
Brink, 198661 24 CT Poor control 14.8 13.3 30
Levy-Marchal, 198862 7 CT Trial 9.3 7.3 43 (3 pts)
Steindel, 1995105 6 CT Non-compliance and 9.2 9.0 –brittle diabetes
Maniatis, 2001106 56, aged Unclear Poor control, flexibility 8.5 8.3 –7–23 of lifestyle, avoiding MDI
Bougneres, 198464 6, aged CT Poor control 11.7 9.3 –1–4 years
Conrad, 2002107 65 MDI Various, including poor 8.4 No significant –control, patient differencepreference, hypos, dawn phenomenon
Boland, 199998 75, allowed CT Post-DCCT research 8.3 on Almost all choice study MDI; 7.5 on continued CSII
CSII
Ahern, 2002108 161, aged CT? but Offered to those 7.1 in 6.5 98% remained18 months– with some interested pre- on CSII18 years intensive schoolers
features such 7.8 in 7.3as BG 7–11 year-monitoring olds
8.1 in 8.112–18 year-olds
Health Technology Assessment 2004; Vol. 8: No. 43
39
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
TABLE 18 Quality assessment of studies comparing analogue versus soluble insulin
Study Random Allocation Group Eligibility Point ITT Withdrawalsconcealment similarity criteria estimates
Bode, 2002113 Adequate Adequate Reported Adequate Adequate Inadequate PartialMelki, 1998116 Unknown Unknown Unknown Adequate Adequate Inadequate InadequateRaskin, 2001117 Unknown Unknown Reported Adequate Adequate Adequate InadequateRenner, 1999114 Unknown Unknown Unknown Adequate Adequate Inadequate UnknownSchmauss, 1998115 Unknown Unknown Unknown Adequate Partial Inadequate AdequateZinman 1997118 Unknown Unknown Unknown Adequate Adequate Inadequate Adequate
See Appendix 3 for description of coding.
TABLE 19 Glycated haemoglobin with analogue and soluble insulin
Study Mean HbA1c (SD)
Analogue Soluble
Bode, 2002113
Parallel RCTLispro N: 28; end of study: 27Regular N: 59; end of study: 50Aspart N: 59; end of study: 55Baseline lispro: 7.3% (0.7).Baseline aspart: 7.3% (0.7).Baseline soluble: 7.5% (0.8)
LisproMonth 4: mean change from baseline0.18% (0.84)
AspartMonth 4: mean change from baseline0.00% (0.51)
SolubleMonth 4: mean change from baseline 0.15%(0.63)
(mean changes from baseline notsignificantly different between the threegroups)
Melki, 1998116
Random crossoverN: 39; end of study: 38Baseline: 7.84% (0.75)Baseline analogue: 7.74% (1.23)Baseline soluble: 7.97% (0.8)
Month 3: 7.11% (SD 0.92)
Change from baseline –0.62 (0.8)
7.88% (SD 0.99)
Change from baseline –0.09 (0.92)Lispro vs regular p = 0.01
Raskin, 2001117
Random crossoverN: 59; end of study: 58Baseline analogue: 7.9% (1.1)Baseline soluble: 7.6% (0.8),p = 0.234
Month 3: 7.41% (SD 0.97)
Change from baseline –0.34 (0.59)
7.65% (SD 0.85), p = 0.004
Change from baseline –0.09 (0.63), lispro vsregular p = 0.004
Renner, 1999114
Random crossoverN: 113Baseline: 7.24% (1)
Month 4: 6.77% (SD 0.88) 6.9% (SD 0.97), p < 0.02
Schmauss, 1998115
Random crossoverN: 11Baseline analogue: 6.3% (0.7)Baseline soluble: 6.7% (1.3),p = ns
Month 3: 6% (SD 0.99)(Mean HbA1c estimated by reviewer)
6.35% (SD 0.83), p = ns
Zinman, 1997118
Random crossoverN: 30Baseline: 8.03% (0.71)
Month 3: 7.66% (SD 0.71) 8% (SD 0.88), p = 0.0041
Clinical effectiveness
40
01 Final valuesMeiki, 1998Raskin, 2001Renner, 1999Schmauss, 1998Zinman, 1997
Subtotal (95% CI)Test for heterogeneity χ2 = 6.65, df = 4, p = 0.16Test for overall effect z = 2.88, p = 0.004
02 Change from baselineBode, 2002
Subtotal (95% CI)Test for heterogeneity χ2 = 0.00, df = 0Test for overall effect z = 0.16, p = 0.9
Subtotal (95% CI)Test for heterogeneity χ2 = 9.09, df = 5, p = 0.11Test for overall effect z = 2.48, p = 0.01
38 7.11 (0.92) 38 7.88 (0.99) 58 7.41 (0.97) 56 7.65 (0.85)113 6.77 (0.88) 113 6.90 (0.97) 11 6.00 (0.99) 11 6.35 (0.83) 30 7.66 (0.71) 30 8.00 (0.88)250 250
27 0.18 (0.84) 50 0.15 (0.63) 27 50
277 300
14.7 –0.77 (–1.20 to –0.34)19.6 –0.24 (–0.57 to 0.09)25.6 –0.13 (–0.37 to 0.11)
6.3 –0.35 (–1.11 to 0.41)15.8 –0.34 (–0.74 to 0.06)82.0 –0.32 (–0.55 to –0.10)
18.0 0.03 (–0.33 to 0.39)18.0 0.03 (–0.33 to 0.39)
100.0 –0.26 (–0.47 to –0.06)
Study mean (SD)Lispro
n mean (SD)Soluble
nWMD
(95% CI Random)Weight
%WMD(95% CI Random)
–1 –0.5 0.5 10Favours Lispro Favours soluble
FIGURE 11 Meta-analysis of the effect of lispro versus soluble insulin on glycated haemoglobin in Type 1 diabetes. Note: subgroup 1‘final values’ include studies reporting mean HbA1c at crossover or end of study (3 months with treatment). Subgroup 2 ‘change frombaseline’ includes one study reporting mean change in baseline HbA1c at end of study (4 months with treatment).
TABLE 20 Insulin dose with analogue and soluble insulin
Study Mean insulin (SD)
Analogue Soluble
Melki, 1998116 Month 3: 0.53 (0.12) U/kg/day 0.55 (0.12) U/kg/day, p = nsRandom crossoverN: 39; end of study: 38Baseline: 0.57 U/kg/day (0.12)
Raskin, 2001117 Month 3: Bolus 0.28 U/kg Basal 0.34 U/kgRandom crossover Basal 0.35 U/kg Bolus 0.30 U/kgN: 59; end of study: 58Baseline not reported
Renner, 1999114 Month 4: 83.1 U/day 85.4 U/day, p = nsRandom crossoverN: 113Baseline: 8.34 U/day
Schmauss, 1998115 Month 3: Basal 19 (6.6) U/day Basal 20 (3.3) U/day, p = nsRandom crossover Bolus 1.4 (0.3) IU/12 g carbohydrate Bolus 1.5 (0.3) U/12 g carbohydrate, p = nsN: 11Baseline not reported
Zinman, 1997118 Month 3: 40.4 (9.3) U/day 40.8 (8.8) U/day, p = nsRandom crossoverN: 30Baseline not reported
difference between analogue and soluble insulingroups. Daily insulin dose for analogue and solubleinsulin varied between 40.4118 and 83.1 U/day114
and between 40.8118 and 85.4 U/day,114
respectively, and between 0.53116 and 0.62 U/kg117
and between 0.55116 and 0.65 U/kg,117 respectively.
WeightThree studies116–118 presented data on weight (Table21). No significant difference in weight gain116 orweight at the end of the treatment period117,118 wasfound between analogue and soluble insulin.
Patient preference and quality of lifeQoL was not reported in any of the studiescomparing analogue and soluble insulin.
Three studies reported patient preference. The majority of patients preferred treatment with analogue insulin than soluble, with 95–100% choosing to continue their treatmentwith lispro, generally owing to its greaterflexibility115,116 (Table 22). Although Schmauss and colleagues115 reported that there was nosignificant difference in treatment satisfaction(method not stated), Renner and colleagues114
found a statistically significant result in favour oflispro with the validated Diabetes TreatmentSatisfaction Questionnaire (DTSQ).
Adverse eventsSevere hypoglycaemic events occurred in 0–8% ofpatients using analogue insulin and 0–10% of
Health Technology Assessment 2004; Vol. 8: No. 43
41
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
TABLE 21 Weight change with analogue and soluble insulin
Study Mean weight (SD) (kg)
Analogue Soluble
Melki, 1998116 Month 3: gain 0.04 (1.79) Gain 0.48 (SD 1.6), p = nsRandom crossoverN: 39; end of study: 38Baseline: 24.4 kg/m2 (2.5)
Raskin, 2001117 Month 3: 79.2 (17.1) 78.8 kg (17.3), p = 0.78Random crossover (1 kg gained at end of period 1, (1 kg gained at end of period 1, N:59, end of study: 58 vs baseline p = ns) vs baseline p = ns)Baseline analogue: 78.3 kg (17.9)Baseline soluble: 77.3 kg (16.7)
Zinman, 1997118 Month 3: 72.6 (9.9) 72.8 (9.9), p = nsRandom crossoverN: 30Baseline: 72.8 kg (9.9)
TABLE 22 Patient preference with analogue and soluble insulin
Study details Analogue Soluble
Melki, 1998116 With which insulin:Random crossover Do you feel better? 34 (89.4%) 2 (5.3%)N: 39; end of study: 38 Are daily activities easier? 32 (84.2%) 2 (5.3%)
Do you prefer? 35 (92.1%) 2 (5.3%)Would you use in future? 36 (94.7%) 2 (5.3%)Gives most flexibility eating at home? 33 (86.8%) 2 (5.3%)Gives most flexibility eating outside? 32 (84.2%) 3 (7.9%)Best balance of glycaemia during study? 34 (89.4%) 2 (5.3%)
Renner, 1999114 Diabetes Treatment Satisfaction Diabetes Treatment Satisfaction QuestionnaireRandom crossover Questionnaire (total score possible 48): (total score possible 48): 32.36 (5.87), p < 0.001N: 113 35.16 (4.25)
Schmauss, 1998115 Continue with treatment: 11 (100%) Continue with treatment: 0Random crossover (owing to greater flexibility with lispro)N: 11 Reports that no significant difference in
treatment satisfaction was noted
patients using soluble insulin, with little differenceapparent between the groups113,115–118 (Table 23).Bode and colleagues113 observed a significantlylower rate of hypoglycaemic events according to symptoms (but not necessarily confirmed byblood glucose <50 mg/dl) in patients in the aspartgroup (6.7 episodes/patient/30 days, SD 5.4)compared with lispro (10.5 episodes/patient/30 days, SD 8.1, p = 0.044) or soluble (10.5episodes/patient/30 days, SD 8.9, p = 0.034)groups. Nocturnal hypoglycaemia with bloodglucose <50 mg/dl was also significantly lower inthe aspart group than the soluble group.113 Melkiand colleagues116 found fewer episodes of bloodglucose <2.0 mmol/l with lispro [0.05 (SD 0.31)per month] than soluble [0.47 (SD 1.17) permonth, p < 0.05], but not of episodes of bloodglucose <3.0 mmol/l. Other studies found nosignificant difference in the number ofhypoglycaemic events defined by patientsymptoms or blood glucose <3 mmol/l,117 meannumber of episodes per patient defined bysymptoms or blood glucose <3.5 mmol/l,114
episodes per 30 days defined by blood glucose<3.5 mmol/l115 or episodes per 30 days defined bysymptoms or blood glucose <3.0 mmol/l.118
No episodes of DKA were reported in four studies.Ketosis was reported in two studies, occurring in1.7–4.4% of patients in the lispro group and0–3.5% of patients in the soluble group.114,117 Nosignificant differences in hyperglycaemia (bloodglucose >350 mg/dl) between aspart, lispro andsoluble insulin were reported by Bode andcolleagues.113 A similar proportion (18–23%) ofpatients in each group experienced catheterobstructions.114,116 In addition, Raskin andcolleagues117 reported that eight (14%) patientsusing lispro and 12 (20%) patients using solubleinsulin experienced 16 and 23 episodes ofhyperglycaemia, respectively, caused byocclusion.117
SummaryAnalogue insulin was associated with a 0.26%lower glycated haemoglobin levels than solubleinsulin and was preferred by patients. Nodifference in insulin dose or weight change wasobserved. Some studies found fewerhypoglycaemic events with analogue insulin,although this varied according to the definitionsused.
Discontinuation ratesOne way of assessing acceptability to patients is toexamine how many patients continue to use
pumps. However, there is a wide range ofdiscontinuation rates and the studies need to beinterpreted with care. Schifferdecker andcolleagues120 noted that discontinuation rates werehigher in studies from the USA, and higher inpatients with poorer control (by HbA1) at the start.They suggest that those with poorer control wereless likely to have been experienced in bloodglucose self-control and insulin dose adjustment,and that discontinuation is less likely if patientshave had a spell on MDI first. This is supportedby Wredling and colleagues,121 who recommendthat patients be selected and prepared for CSII bymeans of at least 6 months of MDI, adequateexperience with blood glucose self-testing, andgood compliance with diabetes management.
Schifferdecker and colleagues also note acorrelation between the discontinuation rate andthe proportion of people with Type 1 diabetes onpumps, with a low rate in European countries suchas Germany, where only 2% of Type 1 diabetespatients were on CSII,120 but a high rate in clinicswith up to 30% of Type 1 diabetes patients onpumps.122 In the study by Knight and colleagues,122
all patients attending the clinic were offered CSII orMDI, and most pump users (101 of 116) wentstraight from conventional insulin treatment withone or two injections per day to CSII. With thisunselective approach, there was a high uptake(30%) of CSII but also a high early discontinuationrate by 12 months, with most of these stoppingwithin 2 months. However, the pump used was theGraseby, and the authors comment that there wasreluctance amongst patients to ‘walking around witha large box attached to me all the time’. Thecommonest reasons for early discontinuation werepump size and discomfort. Another reason washypoglycaemia, in patients who had been poorlycontrolled before CSII, had rarely experiencedhypoglycaemia, and who found occasionalhypoglycaemic episodes unacceptable.
Floyd and colleagues123 reported that 33 (49%) of68 adults starting on CSII stopped, about halfwithin 6 months. A variety of pumps were used,including larger older ones. Factors predictingdiscontinuation included a high baseline HbA1
and a low estimation by users of their ability toself-manage diabetes.
Bell and colleagues124 found in 1988 that one ofthe major reasons (40%) for discontinuation wasneedle site abscess, but in a later paper (1993)125
they report that the cause of the infections hadbeen found to be the diluting fluid, and that therewere no discontinuations due to abscesses in thesecond study.
Clinical effectiveness
42
Health Technology Assessment 2004; Vol. 8: No. 43
43
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
TABLE 23 Adverse events with analogue and soluble insulin
Study Adverse events
Analogue Soluble
Bode, 2002113
Parallel RCTLispro N: 28; end of study: 27Soluble N: 59; end of study: 50Aspart N: 59; end of study: 55
LisproSevere hypos: 0DKA: 0Hyperglycaemia (BG >350 mg/dl): 10, p = ns
≤ 3 clogs or blockages: 64%Hypos [episodes, rate/pt/30days (SD)]:All reported: 872, 10.5 (8.1) vs aspartp = 0.044.Hypos BG <50 mg/dl: 359, 4.4 (4.7) vsaspart p = 0.841Nocturnal hypos BG < 50 mg/dl: 64, 0.6(0.61) vs aspart p = 0.189
AspartSevere hypos: 0DKA: 0Hyperglycaemia (BG >350 mg/dl): 16, p = ns
3 clogs or blockages: 75%Herpes zoster: 1Hypos [episodes, rate/pt/30 days (SD)]:All reported: 1126, 6.7 (5.4)Hypos BG <50 mg/dl: 610, 3.7 (3.6)Nocturnal hypos BG <50 mg/dl: 96, 0.5(0.83)
Severe hypos: 1 (1.7%)DKA: 0Hyperglycaemia (BG > 350 mg/dl): 24, p = ns
≤ 3 clogs or blockages: 78%Hypos [episodes, rate/pt/30days (SD)]:All reported: 1663, 10.5 (8.9) vs aspartp = 0.034Hypos BG <50 mg/dl: 770, 4.8 (4.2) vsaspart p = 0.175.Nocturnal hypos BG <50 mg/dl: 207, 0.9(0.97) vs aspart p = 0.004
Melki, 1998116
Random crossoverN: 39; end of study: 38
Catheter obstructions: 9Severe hypos: 3 (7.9%)DKA: 0Insulin precipitation in catheter: 1
BG <3.0 mmol/l: 7.03 (5.79) per monthBG <2.0 mmol/l: 0.05 (0.31) per month
Catheter obstructions: 9 Severe hypos: 4 (10.5%) patients(7 episodes)DKA: 0Insulin precipitation in catheter: 4
BG <3.0 mmol/l: 7.94 (5.42) per month,p = nsBG <2.0 mmol/l: 0.47 (1.17) per month,p < 0.050 hypos resulted in coma or seizures
Raskin, 2001117
Random crossoverN: 59; end of study: 58
Severe hypos: 3 (5.1%) patients (3 episodes)Ketosis: 1Hyper due to occlusion: 8 pts (16 episodes).
Hypos as defined: 7 pts (8 episodes)
Severe hypos: 2 (3.4%) patients (3 episodes)DKA: 0Hyper due to occlusion: 12 pts (23episodes)[During study 38 pts reported 109 episodesof hyper: 39 caused by occlusion, 47 causedby other reasons, 23 no identifiable cause.Groups not specified]
Hypos as defined: 7 pts (11 episodes).1 hospitalised: fever, vomiting, dehydration
Renner, 1999114
Random crossoverN: 113
Injection site infection: 4
Serious adverse event not related to drug: 1
Hypos: mean 12.4 (13.9), median 8episodes/ptInfection (cold): 19.4%Rhinitis: 15.8%Catheter occlusion: 20 (42 episodes)Ketosis: 5
Injection site infection: 2
Serious adverse event not related to drug: 6
Hypos: mean 11.0 (11.2), median8 episodes/pt, p = nsInfection (cold): 21.1%Rhinitis: 13.8%Catheter occlusion: 21 (45 episodes), p = nsKetosis: 4
continued
Clinical effectiveness
44
TABLE 23 Adverse events with analogue and soluble insulin (cont’d)
Study Adverse events
Analogue Soluble
Schmauss, 1998115
Random crossoverN: 11
Severe hypos: 0DKA: 0Hypos per 30 days: 4 (2.98)
Note: no severe adverse events registeredby either treatment
Severe hypos: 0DKA: 0Hypos per 30 days: 3.2 (2.3), p = ns
Zinman, 1997118
Random crossoverN: 30
Skin reaction: 16Severe hypos: 0
Discoloration of insulin in reservoir orcatheter (did not lead to obstruction orsignificant hyperglycaemia): 2
Hypos: 8.6 (7.7)/30 days (vs baselinep = 0.035)Hypos confirmed by BG: 6.0 (4.9)/30 days(vs baseline p = 0.03)
Skin reaction: 15Severe hypos: 0
Discoloration of insulin in reservoir orcatheter (did not lead to obstruction orsignificant hyperglycaemia): 2
Hypos 10.8 (9.9)/30 days (vs baseline andlispro p = ns)Hypos confirmed by BG: 7.6 (7.1)/30 days(vs baseline p = ns, vs lispro p = ns)
The purpose of this chapter is to give anappreciation of what it means to have Type 1
diabetes that is not well controlled on insulin byinjections and how CSII can change that.Although this chapter gives us access to a widerrange of data than is available from the trials, it isnot methodologically rigorous and findings maynot be generalisable.
CaveatsThe patient’s perspective section is based largelyon written statements from pump users, andseveral caveats are required. First, most commentscome from people who are members of INPUTwho have responded to a request for comments.They are likely to be a more highly motivatedgroup than average and some are clearly highlyorganised individuals. This does not affect thevalidity of their comments, but may haveimplications for generalisability. Second, they aresuccessful pump users and tend to be enthusiastsfor the technology. That is less important becausethose who do not succeed will not incur the on-going costs of pumps. Third, most have had topay for the pumps and consumables themselves,and this creates another selection bias. Fourth,because pumps are little used in the UK, itappears that most of those who have gone on toCSII have done so because they have had a lot oftrouble with control of blood glucose or frequenthypoglycaemic episodes, that is, a severity bias.They may have more to gain than the averageperson with insulin-treated diabetes. It isapparent, for example, that some of theimprovements in HbA1c reported by the INPUTrespondents are greater than seen in the trials.Again, this does not affect the validity of thefindings, but will be relevant to discussions aboutthe proportion of people with Type 1 diabetes whoshould be considered for CSII. It may also meanthat CSII might be more cost-effective in ‘real-lifepractice’ than would be expected from HbA1c
differences seen in the trials.
The term ‘control’ is used in different ways. Theusual use refers to control of blood glucose levels,but we need to distinguish that fairly narrow usagefrom control of diabetes. The broader use includes
control of symptoms, hypoglycaemic episodes andother metabolic disturbances such as highcholesterol levels. Or as one of the INPUTmembers said, “Good diabetic control does notequal normal life.”
Most of the published studies have used outcomemeasures such as HbA1c. Some have includedfrequency and severity of hypoglycaemic episodes,but only one has included QoL.
It may be useful to think in terms of a spectrum ofbenefits in terms of ease of measurement, with themore easily measured such as HbA1c at one end andbenefits such as greater flexibility of lifestyle at theother. Published research tends to focus on themore easily measurable end of the spectrum,whereas it may be that the greatest amount ofbenefit lies at the less easily measurable, or perhapsjust less frequently studied, end of the spectrum.
Since the published studies did not providesufficient data on all outcomes of importance topatients, we have sought information from pumpusers, sometimes directly from individual users,but more from the pump users’ group INPUT.This information has usually come from the user,but sometimes has come from their spouse orparent. For example, few of the trials haveincluded children, and so we have specificallysought advice from a number of parents.
Such information may be from successful pumpusers, but when considering the resource impact,that is appropriate, since those who do not like CSIIwill probably cease to use it and will thereby ceaseto incur the extra costs. Hence the economicanalysis differs from, for example, that of a druggiven to a number of patients, some of whom willbenefit and some not. Those who do not benefitfrom pumps will probably stop using them fairlyquickly. The cost-effectiveness therefore depends onthe costs and benefits for successful long-term users.
Space does not permit the inclusion of commentsfrom all those members of INPUT whoresponded. In the quotations below, they andother pump users are identified only as ‘pumpuser (PU)’ and number. Parents are identified as,for example, ‘Parent 1’.
Health Technology Assessment 2004; Vol. 8: No. 43
45
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Chapter 4
The patient’s perspective
Reasons for switching to insulin bypumpThe reasons for starting CSII given in publishedstudies have been reviewed by Pickup and Keen39
and range from improving poor control to patientpreference.
The meaning of ‘poor control’ and the emphasison HbA1c have been mentioned above. However,as one of our panel of experts noted:
“HbA1c is only an average of what the patient’s bloodglucose has been. It does not show the variations ofblood glucose that any person treated with insulin willexperience. These variations can often be quite wide.An HbA1 measurement cannot show whether thepatient has frequent hypoglycaemia, which is onereason for a person to opt for CSII therapy.” (PU1)
Because HbA1 is only an average, it could reflecteither perfect control with blood glucose alwaysunder tight control, or a situation with both highand low blood glucose levels. As our adviser said:
“From my own personal experience as a pump user of20 months, if only HbA1 was used as the measure ofcontrol, one would not be able to see how CSIItherapy has altered my diabetes control. I managed tomaintain a perfect, i.e. non-diabetes level using MDItreatment for a good 10 years before changing toCSII, and my HbA1 level has not altered with pumptherapy. However, CSII has had a dramatic effect onmy control by reducing the frequency ofhypoglycaemia.” (PU1)
The value of HbA1c is not in doubt as an indicatorof risk of long-term complications of diabetes suchas eye or kidney disease. It is less useful for large-vessel disease such as ischaemic heart disease.
The reasons given by pump users for switching toCSII included:
� To reduce problems with hypoglycaemic events;this was especially the case amongst those whohad lost awareness of developinghypoglycaemia.
� To control hyperglycaemia and improve HbA1c,in order to prevent longer term complications.
� To give control of wide fluctuations in glucoselevels (in effect, the combination of the previoustwo reasons).
� To allow more flexibility in diet and insulin,particularly for those whose schedule may varyfrom day to day or be unpredictable (such asbusiness people, junior doctors andschoolchildren).
� To ‘regain a more normal lifestyle’.� Less painful than multiple injections.� To lose weight, which had been difficult on MDI
because hypoglycaemic events were avoided byregular eating plus snacks when symptoms ofhypoglycaemia occurred, or blood glucosetesting showed lowish results.
� To control the ‘dawn phenomenon’, whichoccurs when the effect of insulin taken theevening before wears off by morning, allowingblood glucose levels to rise to high levels.
� To ensure very good control of blood glucoselevels during pregnancy. (Tight control beforeconception and during pregnancy reduces therisk of malformation and other adverse events.)
One mother reported that her son “likes gadgetsand computers” and wanted to administer hisinsulin with one.
Hypoglycaemic eventsThis was one of the main reasons for switchingfrom MDI to pumps, and a number of respondentscommented on the frequency of hypoglycaemicevents on MDI and the unpredictability of theeffect of long-acting insulin:
“My quality of life has definitely improved mainly dueto the reduction in hypoglycaemia. One very importantimprovement is that my hypoglycaemic warning signsare now much more evident, allowing me to takeappropriate action in reducing the risk of severehypoglycaemia. On MDI therapy, I had problemscontrolling my post-prandial blood glucose even whenusing the quick acting lispro insulin. The choice waseither to take less insulin at meal times to avoid goinghypoglycaemic several hours later, which resulted inpost-prandial highs, or to take more insulin at mealtimes to prevent the post-prandial highs but then riskgoing hypo later and having to eat more food whetherI was hungry or not. The pump solves this problemwith the use of the different boluses available and withthe continuous basal infusion.” (PU1)
“On MDI therapy, I averaged approximately 3admissions yearly to Accident and Emergency byambulance for severe hypoglycaemia. My wife had toadminister glucagon to me on average twice a month;I have only needed one glucagon injection in20 months since being on the pump.” (PU1)
A glucagon kit, with drug, syringe and sterile waterfor injection, costs about £20.126 Hence twice-monthly glucagon would cost about £480 per year.
“At the age of 14 I began to use a basal-bolusinjection regimen. . . . I remember battling exhaustion
The patient’s perspective
46
at school due to fatigue as the result of sleepdisturbance due to nocturnal hypoglycaemia. . . . Ifound that however hard I tried, if I aimed fornormoglycaemia, I would experience frequent hypos,which left me feeling exhausted. . . . Over the next 3years control improved again but I still suffered fromfrequent nocturnal hypos during which I woke upbathed in sweat. I found that reducing long-actinginsulin by 2 units often had little effect and any largerreduction often resulted in morning hyperglycaemia.During my first year employed as a doctor I struggledwith frequent hypos. . . . [On pump therapy] I stillexperience a few hypos but not to the extent of mypre pump days. . . . Overall the pump has enabled meto regain a degree of control over the condition.”(PU2)
“I really do feel that for the first time in my life that Iam actually controlling my diabetes rather than justletting it run constantly high so that I didn’t go hypo.My hypos have become much milder and so mucheasier to control, so that I am not frightened to runmy blood glucose in the normal range.” (PU3)
“I changed to four daily injections for three years.The new regime did not help night time hypos ordawn phenomenon. I experienced severe hypos whichended up with either ambulance call-outs oremergency visits to A and E (12 in two years). . . . Thepump corrected my hypos in the middle of the nightand also my dawn phenomenon. Over a few monthsmy hypo awareness returned in that I would knowwhen my sugar levels started to drop and be able totake glucose without becoming unconscious as before.I have now been using a pump for 6 years and havenot been admitted to A&E since starting on thepump.” (PU4)
Several users commented that awareness ofimpending hypoglycaemia (see ‘Introduction’ forexplanation) had returned once they had been onCSII for a couple of months, probably becausethey had had periods of relative freedom fromhypoglycaemic episodes.
Control of blood glucoseA number of respondents commented that MDIdid not provide satisfactory control.
“I was diagnosed age 12. Years ago I went on to amultiple injection regime, which used long-actinginsulin as a background dose, with injections of short-acting insulin before each meal – a total of sixinjections a day. My diabetes has always been difficultto control. Six injections a day and very regular bloodtesting (5–7 tests a day) backed up by sensible andregular eating and hardly any lifestyle ‘excesses’ didnot enable me to achieve reasonable blood sugarcontrol. Every single day I would have a problem with
either high or low blood sugars. On pump therapyalthough my sugars still fluctuate, on most days I canhope to get through the day without a significantproblem.” (PU 5)
“My diabetes control had been erratic for some yearsdue to problems with insulin absorption after 28 yearsof injections. After I had used the pump for 4months, blood test results decreased by some 20%,which if sustained should lead to decreased morbidityand risk of complications. . . .” (PU6)
“When I was on MDI blood glucoses were erratic. Inone day I could easily go from being hypo to beingover 20. On the pump, I have very few highs andlows. When I do, I can explain each and every one. Ican look back and see that I have misjudged a mealbolus, or that I didn’t reduce my basal rate enoughwhile exercising. Diabetes is a constant balancingact. . . . because of this you have to make small changesto your diabetes control all the time and with pumptherapy this is very simple and very easy.” (PU7)
“My HbA1c results, which date back 4 years beforestarting pump therapy, were at best 7.4% and at worst11.8%, with an average of 8.9%. Within 3 months ofcommencing pump therapy, I achieved 6.8%, and onmy last test it was down to 6.1% – excellent control.”(PU8)
Some respondents had additional comments onthe difference between MDI and CSII:
“To a certain extent some of these things could bepartially accomplished using lispro insulin on MDItherapy. However, the difficulty lies with the injectionof intermediate/long acting basal insulin. For me, Ifound that its action from day to day would peak atimprecise times probably due to the variableabsorption of the unused injected insulin. It is alsoimpossible with this insulin to adjust for changes ininsulin needs and sensitivity throughout the day, e.g.dawn phenomenon. For myself with the pump, I setfour different basal rates throughout a 24 hour day,ranging from 0.1 units/hour for most of my workingday, to 0.4, 0.5 and 0.6 units for the evening, nightand early morning, respectively. This would be verydifficult to do with MDI therapy.” (PU1).
Flexibility of lifestyleThe advantages of pumps in allowing greaterflexibility of lifestyle and working patterns werereferred to by a number of respondents:
“Living with diabetes is not an easy situation.Anything that allows the individual person to takecontrol and manage a medical condition must bebeneficial both to the patient and the carers. Theproper use of an insulin pump allows this to happen.
Health Technology Assessment 2004; Vol. 8: No. 43
47
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
From my own perspective the pump has allowed meto lead a full and active life where I control mydiabetes rather than the diabetes control me. I havebeen able to travel extensively on business and forpleasure without worrying about changing time zones,strange local eating customs and where/when the nextmeal might come from.” (PU9)
“Freedom, flexibility, pleasure and peace of mind inone’s daily life, almost like being a non-diabetic,compared with the uncertainty of the MDI regime.”(PU10).
“I can live a much more flexible lifestyle where I havethe freedom to eat when and what I please withoutcompromising on my glycaemic control, whereasbefore, on MDI, the insulin was controlling my eatinghabits. Now I can more easily regulate the insulin bytaking whatever is needed according to when andwhat I eat. I can even skip meals or take a secondhelping.” (PU1)
“As I have a very busy life the remote control of mypump helps me to be able to eat on the go and give acouple of units of insulin for the carbohydrate that Ieat.” (PU3)
“I have experience of both injection (19 years) andinsulin pump (6 years) therapy. I find pump therapyto be preferable as it gives me far more control of myinsulin input and daily activities. I am now able to livea near normal lifestyle with better control of mydisease.” (PU4).
“I have just completed the first 6 months of anintensive degree course, which has meant balancingfamily life with controlling my blood sugars andworking literally all hours in order to complete myprojects in time. My last three projects involved myworking through the night. . . . I couldn’t imaginedoing that without the support of a pump.” (PU5; aworking mother with young children).
Other commentsThe respondents mentioned a number of benefitsrelating to their reasons for going on to pumptherapy, but they also gave details of benefits thatthey did not seem to have expected:
“ I feel more energetic and more able to cope withthe demands of work.”
“The pump is extremely predictable; I have neverhad a problem with mine. Using the pen, theabsorption of insulin was erratic and everything wasleft to guesswork. . . . ”
“For the first time in my life I almost feel normal. Iam not constantly looking for drink and food.”
“My injection sites were hard and painful and I wouldhave to alternate these. With the pump I do not havethese.”
“I do not have to carry syringes with me everywhereor inject in restaurants or loos before eating”.
“My condition is much less visible, in that I can takeinsulin without anyone being aware of what I amdoing.”
“On the pump I backpacked through Guatemala,Belize and Honduras for 3 months, scuba diving,climbing volcanoes, doing all sorts of activities for thefirst time. I could apply my pump knowledge to anysituation and come out the other end with good BGs.I could conceal my pump so didn’t get needles outwhere locals may have seen and suspected. It wouldhave been a lot harder on injections and moreworrying and stressful.”
“It all goes back to being in control of your conditionrather than the condition controlling you”.
“The extra weight has been lost.”
DisadvantagesThe commonest reported problem was cost.
Other problems included concern about beingattached to a pump 24 hours per day every day:
“At first I was very apprehensive about this form oftherapy since it would mean being connected to apump for 24 h/7 days a week which I didn’t reallythink was for me,” but after being on the pump,“Always being connected to a device is certainly notan ideal way to live. Its advantages however faroutweigh the inconvenience of wearing it. It can easilyand quickly be removed for taking a shower, bath,and swimming or for sex.”
“It certainly made me more aware of my diabetes. Infact to be honest, I had the feeling that I was adisabled person, something I had never felt before.This feeling of being disabled is no longer present.”
“In the 20 months since using the pump, I haveexperienced on 3–4 occasions interruptions of insulindelivery which were not detected by the pump’ssecurity alarm system, and only later detected bytesting my blood sugar and finding an unexpectedhigh. I later discovered the causes for these as eitherbeing a large bubble forming in the infusion line orthat the infusion site cannula had slipped out. Thishas reinforced the need for frequent monitoring ofmy blood sugar.”
“The only problems that I see with the pumps are thered marks that the infusion sets leave on myabdomen, and that I can’t swim with it.”
The patient’s perspective
48
“Curiously enough, I have never found the actualwearing of the pump to be the slightest problem. Thepump and its tubing are easily disconnected at thesite (this takes about a second) should I wish to bewithout it for a short period. It is small and clips tothe waistband of whatever I am wearing and the shortlength of tubing is simply tucked under my clothes.The only time I have found the wearing of the pumpto be a problem was when I was on holiday, wearing aswimsuit and having to keep the pump in a smallcoolbag so that the sun did not overheat the insulin.”
“The only problem occurred shortly after I startedusing it and was unused to it. One evening Iinadvertently knocked out the cannula from under myskin and went to bed. I then had no insulin deliveryuntil the morning.”
The alarm on the pump will go off if the pumpfails to pump out insulin, but as long as the insulinis being pumped out, whether into the patient orthe bed, the alarm will not go off.
Pump users need to remember that they have noreserves of insulin:
“Anyone contemplating pump therapy needs tounderstand that, because one is using only short-acting insulin, if for any reason the supply of insulinis unexpectedly interrupted, the body has no reserveof long-acting insulin, and so sugars can rise sharply.”
and that they have to monitor their blood sugarlevels:
“Anyone on a pump has to accept that they have tokeep a reasonably close eye on their blood sugars, bydoing regular blood tests during the day. This way,any unexpected fluctuations can be investigated.”
and
“If the pump is to work for you, you need to beprepared to work with it. It relies on goodprogramming by the user and you need to bemotivated in order to get the best out of it, as it willnot solve all your problems without any input fromyou”.
With CSII, there is a needle through the skin allthe time, although the site will be changed everyfew days. There is therefore a small but real risk ofinfection:
“The only disadvantage I have found is the very rareoccurrence of infection at the insertion site. In mycase I have suffered this on four occasions in 20 yearswhen I have needed antibiotics. However, it is easy torecognise how I got infections. It was the direct resultof leaving the infusion line in place for too long. This
of course happens because I have to pay for all theancillary supplies, costing about £1200 per annum, acost that would double if I changed the reservoirs andinfusion lines as often as recommended by thesuppliers.”
Children and CSIIAll the comments above came from adult pumpusers. Fourteen submissions were obtained byINPUT from parents of children (some now olderteenagers) with experience of pumps. The numberof children on pumps in the UK appears to be low[currently over 100 children on pumps (Davis J,INPUT: personal communication, 2002)] and thishas been a problem for some families:
“Because there are currently so few others in the samesituation, we had to follow a steep learning curve forthe first few weeks, making up a number of groundrules for ourselves. We have joined a growing supportgroup of pump users. . . . ” (P1)
Space prohibits more than a few quotations fromthe submissions, but a few examples give anillustration of before and after pump therapy.
Before:
“The two words that best summed up living with thecondition are ‘discipline’ and ‘stress’. Disciplinerelates to the way our lives had to be structuredaround Sam’s diabetes. This included arranginginjections to fit in with meal times (i.e. just beforebreakfast and just before tea, in general), working outa fairly rigid regime of snacks and regular meals,trying to ensure that a young child eats just the rightamount at just the right time, forward planning tomake sure that all trips away from home wereadequately prepared for any eventuality. . . . It hadbeen very frustrating to put so much effort intolooking after him, only to see wide fluctuations in hisblood glucose levels partly due to the uncertainties ininsulin absorption from injections.”
(This child developed diabetes at the age of two,and switched to the pump when he was 7.)
After:
“We can definitely say that it has been beneficial forall of us. The first few weeks were very hard work, butonce the various insulin delivery rates had beenestablished, we quickly settled into a routine.
Meal times are no longer rigid or stressful. Sam caneat what he wants (within reason) whenever he wants– we simply add up the amount of carbohydrate ineach meal, convert this to an insulin amount using
Health Technology Assessment 2004; Vol. 8: No. 43
49
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
our own conversion chart, and program the bolus onthe pump. It did not take Sam very long to work outhow to do this himself and it has made an enormousdifference to his independence and self-esteem.” (P1)
From a child switched to CSII from MDI:
“better control on multiple injections than on two aday (HbA1 8.6%) but even better on pump (7%).Pump gives back part of the quality of life that wasdestroyed when he first became diabetic – he may notbe ‘normal’ or ‘better’, but he can live a more or lessnormal life with a pump.”
This child had severe problems with hypoglycaemicunawareness and hypoglycaemic events, had hadhypoglycaemic event-induced seizures and hadepisodes of unconsciousness at school events:
“He was deeply affected by having collapsed at school,in a public hall, on stage in front of all the otherschools in the town with everyone having seen.” [He isa saxophone player.] “He was not allowed his snack ona school trip and he collapsed hypo on the outing.”
After:
“I no longer have the child alarm in his room in casehe goes hypo in his sleep because I am confident he isnot likely to. Having spoken to other parents whosechild has fitted [i.e. had an epileptic-like convulsiondue to hypoglycaemia], it is an experience you never,ever, forget, but on a pump you are able to sleepnights, and so is your child, confident that he willwake up in the morning.” (P2)
Hypoglycaemic episodes affect not just the childbut the family. From the parent of a child whodeveloped diabetes at 1 year old and is now4 years old:
“I have found it hard to go back to work as I seem tobe on call for him all the time. For example, I willdrop him off at 9.30 and by 11 am they can phoneme up because he has gone low, and I have to go backto the school.” (P3)
Most of the respondents noted improvements inHbA1c, but it should be noted that half thechildren were on conventional insulin treatmentwith two injections a day:
“Her HbA1c levels dropped from 10.6 to 8.2, and hermood and personality dramatically changed – we gotour little girl back again.” (P8).
In some cases, treatment with conventionaltherapy rather than MDI was preferred by thehospital clinic, and one parent switched to MDIherself after studying Internet sites and books:
“In a 12 week period, I listed all Jamie’s readings – 78hypers, 63 hypos and 69 normal readings. Went onmultiple injections – I telephoned the hospital and saidthat I would be starting it, as they had failed to respondto my requests for many weeks, and it would be helpfulif they could tell me how to work out the dosage, butthat even if they did not, we would be doing it.” (P2)
Several commented on reduced emergencyhospital visits:
“. . . Has not been admitted to hospital since going onthe pump. On four injections a day had fouradmissions over 2 years.” (P6)
There were comments on the pros and cons ofMDI versus CSII:
“We wouldn’t consider having 12 injections every 3days when we can have only 1 and the ability tocorrect a hyper at the touch of a button with no moreinjections.” (P8)
“I don’t think multiple injections do anything for achild’s mental state. Their lives would be governed bythe clock. Syringes would always have to be carried onthe person which is not socially acceptable (school,discos, etc.) plus the safe carriage of a vial of insulin.”(P9)
Some of the benefits are social rather than medical:
“Steven feels more like his peers now [on the pump].He can eat with them [instead of going in to dinnerearly] and he can eat the same things as them.” (P10)
“Had to move school due to bullying as other pupilscould not accept Nicola eating at certain times of theday. . . .”
“Diabetes [on pump] is now not obvious to anyone asNicola does not need to eat during class and does nothave so many hypo attacks.” (P9)
Some children had snacks stolen; one had hisblood glucose meter stolen.
Disadvantages mentioned included:
� more daily blood tests than on CT, and moreparental input to collate tests and insulin boluses
� cost� the need for vigilance in case any thing goes
wrong with the pump� misunderstandings by other people – “On a few
occasions, people have thought he was playingwith a Game Boy while he was programming abolus”
� being connected to a device 24 hours a day.
However, most respondents said there were nodisadvantages.
The patient’s perspective
50
Common themesSources of informationThe majority of parents got their informationfrom non-NHS sources, the two main ones beingthe Internet, and Balance, the magazine ofDiabetes UK (formerly the British DiabeticAssociation). Several encountered resistance fromthe NHS, and sometimes wrong information:
“We had to go to a private doctor instead of theNHS.” (P4)
“Every avenue had been explored to try to stabiliseher condition (different insulins and regimes) butnothing worked. The pump was a plea from me to seewhether or not it would help, as the consultant didnot know what else to try. Nurse educator fromcompany put her on the pump; hospital consultantdid not get involved as he has no knowledge of thistherapy. It worked, thankfully.” (P9)
“Consultant was patronising, dismissive andcombative.” (P11).
“Consultant thought pumps were dangerous.” (P2)
Training needsMost parents commented that much support wasneeded in the first few weeks. This was sometimesprovided by the local clinic, but often by nursesfrom the pump companies.
Training needs sometimes varied between parentsand children:
“Started pump therapy after a week of training for meand his father – Jamie of course had already workedout how to use the pump from talking to another childwith one and from reading all the manuals.” (P2).
Jamie started using the pump aged 11 years; helikes computers and other gadgets.
FundingOf the 14 respondents, eight were funding thepump and consumables themselves; one hadsupport from a hospital league of friends; fivewere funded by the NHS, some with little trouble,others after much correspondence with the NHS,MPs and the media. It appears that most healthauthorities do not fund pumps for children. Onehealth authority funded the pump for 2 years thenstopped funding.
Problems with schoolsThere were many problems with schools. Some are not relevant to this review, but the testimoniesare consistent in reporting that schools cope much better with children on pumps than those on injections – although to some extent this is because the children find it easier to lookafter themselves, have fewer hypoglycaemicepisodes, do not need to eat at special times, canmiss meals if necessary and do not need to carryinsulin syringes and vials.
ConclusionsThe evidence above differs in type from that seen in traditional systematic reviews, beinganecdotal and subject to selection biases, but itdoes provide valuable information that is notavailable from the published literature, and is very useful. Data on many of the outcomes could be collected in trials. The problem for thisreview is that such data have not been collected ingood-quality studies from which one can extractutility data to feed into cost per QALYcalculations.
Nevertheless, the submissions have been veryuseful, and provide information that is currentlynot available anywhere else.
Health Technology Assessment 2004; Vol. 8: No. 43
51
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
IntroductionThe economic assessment of CSII versus MDI is anevaluation of differing forms of delivery forintensive insulin therapy, and this type ofcomparison has received little, if any, attention inthe scientific literature in terms of cost-effectiveness analysis. Much of the economicsliterature pertaining to diabetes is related to eithercost-of-illness studies127,128 or the cost-effectivenessof intensive diabetic therapy versus conventionaldiabetic therapy (e.g. DCCT129 UKPDS130). Wehave not identified any studies informing on theeconomic consequences of CSII versus MDI.Therefore, we have to see what can be extrapolatedfrom the studies of the benefits of improvedcontrol, and we begin by outlining some of the keystudies, both trials and epidemiological literature,which have been used to support economic analysisin diabetes, and Type 1 diabetes in particular.
A brief outline of the DCCT,19 UKPDS131 andWisconsin Epidemiologic Study of DiseaseRetinopathy (WESDR)132 studies is given below,and in the section ‘Economic models for Type 1diabetes’ (p. 66) we review the main economicmodelling approaches published to date. Thesestudies are used to illustrate the problems ofcreating an economic model to estimate the cost-effectiveness of CSII.
The section ‘Literature review: economicevaluations and quality of life comparisons’ (p. 54)refers to the methodology and findings of theliterature search to identify economic studiesrelevant to the evaluation of CSII versus MDI. Thesection ‘Costs associated with CSII (versus MDI)’(p. 55) presents the cost analysis for CSII. Thesection ‘The benefits of CSII’ (p. 59) provides anassessment of the benefits and QoL issuesassociated with CSII. The sections ‘Economicmodels for Type 1 diabetes’ (p. 66) and ‘Cost-effectiveness of CSII versus MDI’ (p. 71) explorethe opportunities to estimate the cost-effectivenessof CSII using available models and through thesynthesis of available cost and outcome data.
The Diabetes Control andComplications Trial (DCCT)The DCCT19 was a multi-centre randomised
clinical trial designed to compare the effects ofintensive diabetes therapy with those ofconventional diabetes therapy on the developmentand/or long-term progression of diabetescomplications of IDDM. The intensive therapy wasdesigned to achieve blood glucose values as closeto normal range as possible with three or moredaily insulin injections or treatment with aninsulin pump. Conventional therapy consisted ofone or two insulin injections per day (note thatconventional insulin therapy was probably lessthan in the UK, where most patients would get twoinjections of mixtures per day). Two cohorts ofpatients were studied, a primary preventioncohort, to consider whether intensive therapywould prevent the development of complications,and a secondary cohort, to consider whetherintensive therapy would affect the progression ofearly complications. Retinopathy was the principalstudy outcome, but renal, neurological andcardiovascular outcomes were also studied, as werethe adverse effects of the two regimens. Theresults of the DCCT demonstrated that intensivetreatment led to a significant risk reduction in theonset and progression of retinopathy, nephropathyand neuropathy. However, intensive therapy(compared with conventional therapy) wasassociated with a threefold increase in the risk ofsevere hypoglycaemia.19 Analyses did notdemonstrate differences in the QoL outcomesbetween the treatment groups. Most economicsassessments in the field of diabetes have beenundertaken using largely homogeneous modellingmethods, which utilise the data from the DCCT.
The model developed and presented by theDCCT Research Group129 examined the cost-effectiveness of alternative approaches to themanagement of IDDM. The model was used toconsider all persons with IDDM in the USA whowould meet the DCCT eligibility criteria, in orderto provide estimates of the lifetime benefits andcosts of intensive therapy, and to address whethermore costly intensive therapy would be preferableto conventional therapy from the perspective ofthe healthcare system. The model used datacollected as part of the DCCT, together with datafrom other clinical trials and epidemiologicalstudies – detail on the model is presented later inTable 35. The DCCT Research Group present
Health Technology Assessment 2004; Vol. 8: No. 43
53
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Chapter 5
Economic analysis
findings from the model to demonstrate thatintensive therapy reduces complications and canbe expected to increase length of life. The authorsalso state that from a healthcare systemperspective, intensive therapy is well within therange of cost-effectiveness considered to representgood value129 – although the marginal benefits ofthe intensified treatment will vary depending onthe level of control achieved with whateverconventional treatment is used.
The DCCT data are not appropriate to assess therelative benefits of CSII and MDI, because theintensive treatment group was treated with eitherMDI or CSII with non-randomised allocation.However, the study does provide evidence on therelationship between HbA1c and health outcomes(long-term complications), and on some of theQoL issues related to intensive therapy andhypoglycaemia. The DCCT also offers a generalfoundation for modelling the cost-effectiveness oftherapies in IDDM.
The United Kingdom ProspectiveDiabetes Study (UKPDS)The UKPDS, started in 1977, was designed toestablish whether, in patients with Type 2 diabetes,intensive blood glucose control reduced the risk ofmacro- or microvascular complications, and toinvestigate the benefits of the therapeutic options.Studies within the UKPDS have reported on bloodglucose control with sulphonylureas or insulincompared with conventional therapy,9 control withmetformin on complications in overweightpatients133 and cost-effectiveness analysis for bothof these comparisons,130,134 amongst others. TheUKPDS has reported that intensive blood glucosecontrol in patients with Type 2 diabetes cansubstantially reduce the cost of diabeticcomplications and increase the time free ofcomplications. Findings from the UKPDS supportthose findings presented in the DCCT concerningthe benefits of improved control.
As with the DCCT, the UKPDS presents findingsfrom comparisons of intensive versus conventionaltherapy. For example, the UKPDS9 reports a 0.9%difference in HbA1c between intensive (7.0%) andconventional (7.9%) groups over 10 years (an 11%reduction), and data showed a significant 25%reduction in microvascular endpoints(p = 0.0099), most of which was a risk reduction of21% for retinopathy. The study suggests that therisk reduction is due to improved glycaemiccontrol, rather than the method by which it wasachieved, as there were no significant differencesbetween the three types of drug treatment.
However, the link between glycaemic control andoutcome is complex as the HbA1c levelsprogressively increased over time, although thedifferences between the more intensively treatedand control groups was maintained.9
As part of the UKPDS,130 a model was developedto estimate cost and effect. The model is a discreteevent simulation model, looking at cost per event-free year; however, it does not have a direct link tothe HbA1c input variable.
The Wisconsin Epidemiologic Study ofDiabetic Retinopathy (WESDR)The WESDR is a population-based study of theprevalence and incidence of diabeticretinopathy.129 The study population consisted ofa sample selected from 10,135 patients withdiabetes who received primary care in an 11-county area in southern Wisconsin from 1979 to1980. The study reports on ‘younger onset’ Type 1diabetic patients diagnosed before 30 years of age.The study provides unique long-term population-based information. Within the sample studied the14-year rate of progression of retinopathy was86%, regression of retinopathy was 17%,progression of proliferative retinopathy (PDR) was37% and the incidence of macular oedema was26%. Data from the study suggest relatively high14-year rates of progression of retinopathy andthat a reduction of hyperglycaemia andhypertension may result in a beneficial decrease inthe progression to proliferative retinopathy. Datafrom WESDR have been applied in diseaseprogression models used for the purpose ofeconomic evaluation [e.g. Palmer andcolleagues,135 discussed further in the section‘Economic models for Type 1 diabetes’ (p. 66)].
Literature review: economicevaluations and quality of lifecomparisonsMethodsLiterature searches were undertaken using searchterms related to the technology and to economicsand QoL. Searches were carried out in a range ofdatabases (see Appendix 2).
Summary of findings on cost-effectivenessThere were no studies identified via the literaturesearch that reported on the cost-effectiveness ofCSII versus MDI. Experts were also consulted inorder to identify any literature on the cost-effectiveness of CSII, but no further studies were
Economic analysis
54
identified. Furthermore, we did not identify anyliterature covering the costs associated with CSII.
Summary of findings on quality of lifecomparisons The literature searches provided only one RCTcomparing QoL outcomes of CSII versus MDI.This study79 found no significant differencesbetween the two interventions [see the section‘CSII versus MDI’ (p. 14)]. Some descriptive andobservational studies that assist with theunderstanding of the QoL consequences of CSIIare discussed in the section ‘The benefits of CSII’(p. 59).
Costs associated with CSII (versus MDI)Additional costs associated with CSII have beenestimated based on the perspective of the NHSand personal social services sector. The costanalysis here aims to identify cost differencesbetween CSII and MDI. Costs componentscomprise (i) intervention costs, (ii) differences ininsulin costs, (iii) differences in the generalmanagement of patients, (iv) the costs associatedwith additional education for pump users and (v) costs associated with adverse events andcomplications. From these, we provide an overallestimate of the additional costs associated withCSII. Cost estimates are based on data from a
number of sources (i.e. manufacturers, patientgroups and expert opinion from two diabetescentres with expertise in management of patients on CSII therapy). As previouslymentioned, we found no cost data in thepublished literature.
Intervention costsTable 24 shows the differences in the suppliesrequired for CSII and MDI therapy, indicatingwhere the intervention costs differ.
Insulin pump costsIn the UK there are currently three insulin pumps available for diabetic patients, DisetronicH-Tron, Disetronic D-Tron and MiniMed 508.Table 25 provides the costs associated with thepurchase and maintenance of the insulin pumpsavailable.
One aspect of the cost analysis is the assumptionon the expected life of the insulin pump. Theguaranteed life of those insulin pumps available inthe UK is 4 years, whereas indications fromINPUT and insulin pump manufacturers arearound an expected life of 7–8 years. This analysiswill use as a base case a pump life of 4 years,reflecting the guaranteed life where nomaintenance costs are payable. Further costcomparisons and sensitivity analysis will explorealternative periods of expected pump life (e.g.8 years).
Health Technology Assessment 2004; Vol. 8: No. 43
55
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
TABLE 24 Supplies used in diabetes treatment by intensive MDI therapy or CSII therapy
Supply item Needed for MDI? Needed for CSII? Additional costs for CSII compared with MDI
Lancets Yes, ≥ 3 per day Yes, ≥ 4 per day No
Glucose test strips Yes, ≥ 3 per day Yes, ≥ 4 per day No
Glucometer Yes Yes No
Insulin Yes, short-acting and other Yes, short-acting insulin; unit Reduced insulin use in CSII intermediate- and long-acting requirements vary userspreparations; unit requirements vary
Insulin syringes or insulin Yes, ≥ 3 per day Yes No (minor cost saving in CSII pen users)
Insulin infusion pump No Yes (see pump costs for Yesdetails)
Insulin cartridge/reservoirs No Yes Yes
Adaptors for insulin No Yes, depending on type of Yescartridge/reservoirs insulin pump
Infusion sets No Yes, one set every 3 days Yes
Batteries No Yes, battery life varies by Yespump type
Estimates of the cost for consumable itemsrequired as part of CSII therapy are presented inTable 26. All pumps require batteries, infusion setsand insulin cartridges and/or adaptors.
Insulin costsThe clinical review has shown a reduction in thedaily insulin dose with CSII. Table 27 uses thefindings from the meta-analysis [see the section‘CSII versus MDI’ (p. 14)] to estimate the costimplications of the reduced insulin requirement.
Cost for general patient managementDiabetic patients on MDI are typically managedon an outpatient basis at a local diabetes centre.Patients typically (where no complications occur)attend for a regular annual outpatientappointment with the consultant physician, and at6-monthly intervals they attend for anappointment with a DSN (two visits per year).They should be able to have HbA1c measuredevery 3–4 months. Patients also have access toadvice by telephone to the medical team at their
Economic analysis
56
TABLE 25 Costs associated with purchase and maintenance of insulin pumps
Type of insulin pump Initial cost (£) Expected life Guaranteed pump Maintenance cost: (years) life (years) average cost per year
following warranty life (£)
Disetronic H-Tron 2350 4–8 2 pumps at 2 years each 176.25(total 4)
Disetronic D-Tron 2468 4–8 4 176.25
MiniMed 508 2562 4–8 4 381.88 for 2 years
Costs are inclusive of VAT.Source: INPUT136 and pump manufacturers.
TABLE 26 Cost for consumable items required by insulin pump users
Item Unit cost Requirements Cost per year (excluding VAT) (£) (including VAT) (£)
Insulin cartridge/reservoirs:Disetronic H-Tron 41.25 per 25 Depend on dose 51.47Disetronic D-Tron 35.00 per 10 Depend on dose N/Aa
MiniMed 508 40.80 per 24 Depend on dose 53.60
Adaptors for insulin cartridge/reservoirs:Disetronic D-Tron 35.00 per 10 Weekly (approx.) 220.10
Infusion sets:Disetronic H-Tron and D-Tron Varies: 4.04/4.96/5.25 1 every 2 or 3 days 950.15b
MiniMed 508 Varies: 3.70–7.05 1 every 2 or 3 days 847.60c
Batteries:Disetronic H-Tron 10.30 per set 1 set every 5 weeks 126.21Disetronic D-Tron 11.00 per set 1 set every 12 weeks 56.17MiniMed 508 4.50 per set 1 set every 6 weeks 45.95
Estimated annual cost for consumable items (including VAT):Disetronic H-Tron 1127.83Disetronic D-Tron 1226.42MiniMed 508 947.15
a Disetronic D-Tron is able to use 3-ml Humalog insulin cartridges.b Weighted annual cost from Disetronic/Abacus.c Weighted annual cost based on average cost for Teflon needles and proportions of 0.55 and 0.45 for Teflon and steel
needles used.Source: cost data from INPUT 2001,136 unless stated otherwise.
diabetes centre (normally their DSN). This level ofcare is not expected to be any different once apatient has started CSII therapy (assuming thateducation support is adequate for those on MDI,otherwise there would be an added cost) and hasbeen through the education and training requiredas part of CSII therapy. However, in the earlystages of CSII therapy it would be reasonable toexpect that one extra visit to the DSN may benecessary (in addition to the education required atthe start of CSII therapy), at 3 months after thestart of pump therapy (source: expert opinion)(Table 28). These extra visits probably do not applyin the case of children, who are usually seen moreoften (every 3 months).
Where complications do occur, the treatmentpatterns/care pathways will be the same for bothMDI and CSII patients.
Costs associated with education forCSIIEducation for CSII: patient levelWhere patients are switched from MDI to CSIIthey must undertake a programme of education tofamiliarise themselves with the use of the insulinpump and its use in the management of theirdiabetes. There are no published data on the costsassociated with insulin pump education. Weconsulted Bournemouth Diabetes and Endocrine
Centre, a leading centre for insulin pump use, andtheir practice indicates that all patients on MDIshould have been through diabetes education aspart of the management of their diabetes. Thetypical education package at the BournemouthCentre, for MDI, would entail an educationprogramme comprising attendance at four 6-hoursessions, in a consecutive 4-week programme.These programmes typically cater for groups of5–6 patients and involve a DSN and dietitianthroughout the programme, with a consultantphysician devoting up to 3 hours of contact timeover the complete programme.
Diabetic education will differ across centres fromformal packages of education, as provided inBournemouth, to more informal and ad hocpatient education programmes. The issue in thisreview is the additional education required bypatients when switching to CSII. Again we havedrawn on current practice at the BournemouthCentre, which we believe provides a reasonableestimate (a relatively thorough package ofeducation and training) of the education necessaryfor insulin pump users. When a patient initiatesCSII, additional education is provided at bothgroup and individual patient levels. Patientsattend for both group and individual educationsessions. Education is delivered to groups of3–5 patients (minimum, maximum) and comprisesinput from a DSN, at 9-hour group level, andthereafter patients typically attend for 3 hours ofeducation with the DSN at an individual patientlevel. The typical programme would comprise a 6-hour group session, a 3-hour group session andindividual education (3 hours) thereafter. Inputfrom a dietitian would be variable, from0–6 hours, depending on the level of prioreducation within the group (i.e. MDI educationpackage outlined above) and the time that had
Health Technology Assessment 2004; Vol. 8: No. 43
57
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
TABLE 27 Reduction in insulin dose and associated costs
Reduction in daily insulin dose: Source Insulin/cost per unita Annual cost mean (95% CI) reduction (£)
–9.73 (–14.5 to –4.91) Meta-analysis 2.5–4 months Humalog: 5 × 3-ml cartridge 66.33(all appropriate trials) (100 units per ml) = £26.78
Soluble: Actrapid, 10-ml vial 37.29(100 units per vial) = £10.50
–13.63 (–23.62 to –3.64) Meta-analysis 2.5–4 months Humalog: 5 × 3-ml cartridge 88.72(RCT data only) (100 units per ml) = £26.78
Soluble: Actrapid, 10-ml vial 52.57(100 units per vial) = £10.50
a Source of cost data: BNF, March 2002.
TABLE 28 Additional resources for general patientmanagement: CSII versus MDI
Additional resource items Additional cost (£)
1 additional outpatient 74appointment in year 1
Source: PSSRU, 2001.
lapsed since the delivery of prior educationprogrammes. In the estimate provided in Table 29we assume the maximum expected 6 hours ofdietitian input. In paediatric care, because ofsmaller numbers, group sessions may not bepossible.
Education for CSII: institutional levelStaff in diabetes centres providing CSII therapywill also require additional education and trainingwhere it is lacking. There is an organisation set upin the UK, Pump Management for Professionals(PUMP), a collaboration between BournemouthDiabetes and Endocrine Centre and the HarrogateDiabetes Centre, that delivers education toprofessionals to facilitate the wider use of insulinpump therapy. PUMP recommends that for eachcentre managing patients on insulin pumps aminimum of three persons be educated on CSII:one physician, one dietitian and one DSN. Theeducational programmes offered currentlycomprise a 3-day teaching programme. Table 30presents details of the costs associated with theeducation of diabetes centres.
Information from PUMP indicates that 90 diabetescentres currently have some patients (one or more)on CSII and 25 centres have 10 or more patients.It is difficult to estimate the number of centresthat would require CSII education for health careprofessionals. It may be that some of those centresalready managing CSII patients would still require
some degree of professional education. At one endof the spectrum we could assume that all 218 NHSTrusts in England and Wales that offer diabeticcare, as listed in the Directory of Diabetes Care,137
require the minimum education detailed in Table 30. This would result in an estimate of about£590,000 to cover education as described in theabove table. However, it is unlikely that all centreswould be responsible for the management of CSIIpatients, with the setting up of local or regionalCSII centres being a possibility, which wouldreduce costs.
Costs associated with severe hypoglycaemiaThe clinical review suggests that there will beslightly fewer severe hypoglycaemic events on CSIIthan MDI. Severe hypoglycaemia could have aneconomic impact, owing to the costs associatedwith treatment and the potential impact onpatient QoL. However, indications from the twodiabetes centres assisting with our review(Bournemouth and Harrogate) are that mostinstances of severe hypoglycaemia are treated bypatients and carers and do not result in significantNHS resource implications. There is littleliterature on the cost consequences of severehypoglycaemia. We detail below the insightsavailable from the limited literature and thefindings from consultation with two diabetescentres. We use these in Table 31 to give anestimate of the cost per severe hypoglycaemicevent, based on the assumptions on treatment
Economic analysis
58
TABLE 29 Patient-level education for CSII therapy
Input to education programme Hours Staff inputs: cost per houra (£) Cost per patient (assuming a group of 4 patients) (£)
DSN – group 9 21.75 48.94Dietitian – group 6 22.23 33.35DSN – individual 3 21.75 65.25Additional education cost per patient switching to CSII 147.54
a See Appendix 14 for details.
TABLE 30 CSII – education and training for professionals
Cost/resource inputs Number (minimum) Cost per daya (£) No. of days Cost estimate (£)per centre
Physician 1 578 3 1734DSN 1 163 3 489Dietitian 1 164 3 492Course fees (NHS transfers) N/A N/AEstimate of the costs associated with education for professionals 2715
a See Appendix 14 for details.
given in the third column. Based on discussionwith the diabetes centres the cost profile for severehypoglycaemic events assumes that in most casesthere is no input from NHS staff, with patients orcarers administering glucagon. In a number ofinstances patients are expected to attend or reportthe event to a DSN and we have made anassumption that this occurs in 10% of cases. Also,we have made an assumption that in 5% of casesan ambulance is called out to a patientexperiencing an event, although these cases donot result in a hospital admission. These areconservative assumptions, and in reality there maybe fewer instances of glucagon use, fewervisits/reports to the DSN and fewer ambulancecallouts than those detailed in the cost profile.
The NHS Reference Costs Data139 offer estimatesfor two health-related resource groups (HRGs)(K11, K12) associated with emergencyhypoglycaemia in diabetic patients. K11 refers to‘Diabetes with Hypoglycaemic Emergency >69 orw cc’ and K12 refers to ‘Diabetes withHypoglycaemic Emergency <70 w/o cc’. The costdatabase provides estimates for elective inpatient(ELIP) costs, non-elective inpatient (NELIP) costsand day case (DC) costs. Although electiveinpatients stays (costing £1395 for K11 and £796for K12) do not seem correct for emergencyhypoglycaemia, 48 such events are listed asfinished clinical episodes (FCEs) in the ReferenceCosts Database and 198 for K12. It is possible thatthey have been mis-coded as elective, or thatpatients have been admitted as semi-elective cases in order to review treatment and improvecontrol. For NELIP costs are £954 and £567,respectively (activity data show 3027 and 2416FCEs). For DC cost estimates are £222 and £352,respectively (16 and 97 FCEs). These HRG costsdo not reflect the costs of the average severehypoglycaemic event. It is possible that the costestimates given in the NHS Reference Costs referto severe hypoglycaemia complicated byuncommon related events, possibly hypoglycaemiaassociated with seizure or coma. Althoughrelevant, they are probably sufficiently rare to
have little effect on the average cost of a severehypoglycaemic event.
A Scandinavian study by Nordfelt and Jonsson140
reports a 12-month prospective follow-up ofdiabetic patients and estimates the costs associatedwith diabetes, providing an estimate of the costsassociated with severe hypoglycaemia. The studyassesses costs and other short-term effects ofsevere hypoglycaemia in 129 children andadolescents with Type 1 diabetes. Data werecollected on 111 events of severe hypoglycaemia(16 were events involving unconsciousness and 95were non-unconscious events). The average costper event is reported at €239 (2047 SEK) forevents involving unconsciousness and €63 (543SEK) for events not involving unconsciousness (n= 95). These estimates include indirect costs (e.g.time lost from the workplace), and the estimateswith healthcare costs only are substantially lower at€151 and €28, respectively (approximately £94and £17.50, using an exchange rate of €1.6 per £).The study by Nordfelt and Jonsson, although inchildren and adolescents, adds support to our costassumptions for severe hypoglycaemia (as above).
In a study available in abstract only, Leese andcolleagues from the population-based DARTScollaboration reported that over a 12-monthperiod, 7% of patients with Type 1 diabetes hadsevere hypoglycaemic events (‘severe’ beingdefined as needing emergency assistance fromNHS personnel). One in three episodes (of allcases, Type 1 and Type 2 diabetes) were dealt withsolely by the ambulance crew. The average costwas around £377.141
Estimated marginal cost for CSIIBased on the cost estimates presented above, theadditional costs of CSII are as given in Table 32.
The benefits of CSIIFrom the clinical effectiveness evidence and theusers’ submissions, the benefits of CSII over MDIinclude:
Health Technology Assessment 2004; Vol. 8: No. 43
59
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
TABLE 31 Cost estimate for treatment of severe hypoglycaemic event
Treatment Unit cost (£) (source) Proportion of events Proportionate cost (£)treated in this way (%)
Glucagon 19.95 (BNF126) 100 19.95DSN/outpatient attendance 74.00 (PSSRU138) 10 7.40Paramedic/ambulance attendance 247.00 (PSSRU138) 5 12.35Average cost for severe hypo 39.70
� QoL gains from greater lifestyle flexibility� improved diabetes control as reflected in HbA1c,
which might be expected to reduce long-termcomplications
� a reduction in hypoglycaemic events, whichwould be expected to be associated with areduction in the fear of hypoglycaemic events
� an improvement in hypoglycaemic eventawareness, which will feed into both theprevious benefits.
However, all these benefits can feed into animproved quality and quantity of life.
This section starts with an overview of diabetesand QoL and then considers individual benefits.
Diabetes and quality of lifeA recent review of QoL associated with diabetes142
characterises the nature of diabetes (generally) byreporting that people with diabetes often feelchallenged by their disease and its day-to-day
management demands, finding that thesedemands are substantial. Patients with diabetesmust deal with their diabetes all day, every day,making countless decisions in an often futile effortto approximate the non-diabetic metabolic state.Diabetes therapy, such as taking insulin, cansubstantially affect QoL either positively, byreducing symptoms of high blood sugar, ornegatively, by symptoms of low blood sugar. Thepsychosocial toll of living with diabetes is often aheavy one, and this toll can often affect self-carebehaviour and, ultimately, long-term glycaemiccontrol, the risk of developing long-termcomplications and overall QoL.142
However, to place diabetes in context, the authorsof the above review report that ‘people withdiabetes have a worse QoL than people with nochronic illness, but a better QoL than people withmost other serious chronic diseases’ (p. 205). Thereview is a useful one, albeit general, and presentsfindings on a range of issues affecting the
Economic analysis
60
TABLE 32 Estimated marginal cost associated with CSII compared with MDI
Cost consequence of CSII versus MDI Disetronic D-Tron (£) Disetronic H-Tron (£) MiniMed 508 (£)
Intervention costs:Pump 2,468 2,350 2,562Consumables 1,226.42 1,127.83 947.15
General patient management 74 74 74
Patient education costs 147.54 147.54 147.54
Difference in treatment cost for severe hypos –39.70 –39.70 –39.70(assuming a reduction of 1 event per year)
Difference in insulin costa –88.72 –88.72 –88.72
Total net cost for CSII:
Year 1 3,878 3,571 3,602
Cumulative cost, assuming 4-year pump life:
Years 1–4 (discounted)b 7,081 (6,722) 6,569 (6,242) 6,058 (5,790)
Mean non-discounted cost per year 1,770 1,642 1,514(costs spread over 4 years)
Years 1–8 (discounted)b 13,941 (11,871) 12,917 (11,011) 11,894 (10,201)
Mean non-discounted cost per year 1,743 1,615 1,487(costs spread over 8 years)
Cumulative cost, assuming 8-year pump life:
Years 1–8 (discounted)b 12,178 (10,429) 11,272 (9,663) 10,096 (8,728)
Mean non-discounted cost per year 1,522 1,409 1,262(costs spread over 8 years)
Education for professionals ~2,715 per centre, ~2,715 per centre, ~2,715 per centre, year 1 only year 1 only year 1 only
a Using data from meta-analysis including RCTs alone.b Figures in parentheses indicate where discounting of future costs has been undertaken using a discount rate of 6%.
relationship between QoL and diabetes, forexample, type of diabetes, treatment regimen,glycaemic control, demography and patientcharacteristics. The review highlights some of theassociations and uncertainties evident in theliterature covering the assessment of QoL indiabetic patients generally.
A number of the above issues impact on theassessment of CSII, which may affect glycaemiccontrol, adverse events and lifestyle issuesassociated with managing diabetes. Within thisassessment of CSII versus MDI, the prime concernis the difference between the two therapeuticoptions. With respect to QoL, there are fewinsights as to any potential differences between thetwo options. From the clinical review only onestudy has reported QoL as a trial outcome, by Tsuiand colleagues,79 and that study failed to find asignificant difference between CSII and MDI(using DQOL). Further literature searches did notidentify any studies (randomised or crossoverstudies) comparing QoL outcomes in CSII andMDI. A number of descriptive studies may assist informing a view on the QoL consequences of CSII,and these are discussed below. However, we mustbe aware that the findings from these studies maynot be generalisable and may introduce bias froma variety of sources, such as patient selection andstudy design.
From the clinical review findings in Chapter 3, thesmall, but significant, difference in HbA1c and thereduction in the event rate for severehypoglycaemia are deserving of attention in terms of their impact on the QoL of patients (CSII versus MDI). These issues are discussedbelow.
Glycaemic control (HbA1c)Data from long-term trials and epidemiologicalstudies have provided evidence that ‘good’metabolic control protects diabetic patients againstchronic complications (e.g. DCCT ResearchGroup,19 WESDR132 and UKPDS130).
The DCCT Research Group143 have presentedfindings on the relationship of glycaemic exposure(glycated haemoglobin/HbA1c) to the risk ofdevelopment of progression of retinopathy in theDCCT. The DCCT demonstrated a markedreduction in the risks of development andprogression of retinopathy and othercomplications of IDDM with intensive treatmentcompared with conventional treatment. They alsopresent findings143 from an epidemiologicalassessment of the association between levels of
HbA1c, before and during the DCCT, with the riskof retinopathy progression. The DCCT data donot provide a definitive assessment of the causalrelationship between specific levels of glycaemicexposure (HbA1c) and the risk of complications,143
as confounding is possible from a number ofsources. However, the findings from theepidemiological assessment lead the DCCT groupto believe that the mean HbA1c (i.e. improvedHbA1c) is the dominant predictor of the reducedrisk of complications in the intensive treatmentgroup, and that data on the total exposure toglycaemia (based on HbA1c at screening and thelevel during the trial), the baseline duration ofIDDM and the time in the study provide a strong,although indirect, affirmation of the glucosehypothesis (i.e. lower levels of glycaemia wouldreduce the development and progression ofcomplications).
The DCCT Research Group use a number ofdifferent statistical models (proportional hazards,Poisson and Logit models) to show that the risk ofcomplications increases with increases in themodel variables reflecting glycaemic exposure. It isfrom these sources that the recent meta-analysis byPickup and colleagues,40 comparing CSII withMDI, postulates that a difference in HbA1c of 0.5%approximates to a reduction of 0.5 cases ofsustained retinopathy progression per 100 patientyears (such a difference would be reflected by areduction in the probability of sustainedretinopathy progression of p = 0.005, or 0.5% perpatient per year). Although these statistical modelsare helpful, they do not lend themselves to anapplication in economic models, as the differencein HbA1c cannot be translated into a transitprobability from meaningful health statedescriptions [i.e. a Markov or decision analyticmodel requires a pathway of disease from onehealth state to another and the DCCT statisticalmodels do not offer this opportunity, based on thepublished literature. This point is discussedfurther below, in the section ‘Cost-effectiveness ofCSII versus MDI’ (p. 71), as we assess the generalfindings from the DCCT statistical models thatpredict the absolute risk of onset of sustainedretinopathy progression].
Data from WESDR demonstrated the generalrelationship of glycaemic control to the incidence of diabetic complications.144 WESDRhealth outcomes data suggest that higher levels of glycaemia are related to a decreasingQoL, but the investigators expect this to bebecause of the increased incidence ofcomplications.145
Health Technology Assessment 2004; Vol. 8: No. 43
61
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
HypoglycaemiaHypoglycaemia is the most common complicationof therapy for Type 1 diabetes, and the reportedincidence varies considerably (from four to 65episodes per 100 patient years).146 Theconsequences of hypoglycaemia vary from mildsymptoms and signs to profound sequelae ofneuroglycopenia including coma and seizures.146
Effects on motor coordination, cognitive functionand judgement can impair performance ofcomplex functions such as driving a car, and alsosimple tasks, such as treating hypoglycaemia withoral carbohydrate. Mild hypoglycaemia can berapidly corrected by food intake or sugary drinks,and has little clinical impact. However, severehypoglycaemia, generally defined as requiringassistance from another person, can be serious,involving the potential for harm to patients andothers. Severe hypoglycaemic events may alsoresult in coma or seizure.
Severe hypoglycaemia is common in all patientswith Type 1 diabetes with almost one-third ofpatients experiencing one or more episodes eachyear (EURODIAB IDDM Complications StudyGroup147). Severe hypoglycaemic events arethemselves short-term events, with acute effectslasting no more than 1–2 days in most cases.However, the fear of experiencing a hypoglycaemicevent is a further characteristic of the disease.
The clinical review [see the section ‘Results inadults with Type 1 diabetes’ (p. 16)] found that therate of severe hypoglycaemic episodes differedlittle between treatments in moststudies,79,80,83,85–88,90,91 with no severe episodesoccurring in two studies.86,90 However,significantly fewer severe hypoglycaemic eventsper 100 patients years occurred with CSIIcompared with MDI in the study by Bode andcolleagues.81 Therefore, the impact of thispotential reduction in the rate of severe events onthe QoL of patients may be an important factor inthe comparison of the two treatment options. Asdiscussed in Chapter 3, it is possible that the RCTsunderestimate the benefits of CSII for those withrecurrent severe hypoglycaemia.
Findings from the DCCT showed that althoughhypoglycaemia is an important adverse event forIDDM, the relationship between hypoglycaemiaand QoL is not clear. The DCCT presents findingson the relationship between hypoglycaemia andQoL as measured within the trial. QoL in theDCCT was assessed through the completion of theDQOL measure, the Symptom Checklist-90R(SCL-90R), the Short Form with 36 Items (SF-36)
and psychosocial events observed in the trial. QoLdata were collected from patients at annual visits.Initial analyses did not demonstrate differences inthe QoL of outcomes between the treatmentgroups,19 even though there was a markedincrease in hypoglycaemia in the intensiveintervention group. Subsequent analyses havesupported this finding.148 End of studyassessments showed that the scores on all scalesdid not differ between the treatment groups.Subsequent analyses were undertaken to assess therelationship between hypoglycaemia and QoLoutcomes. The DCCT Research Group148
modelled the relationship of hypoglycaemia to theSCL-90R and DQOL outcomes. Data showed thatthe occurrence of severe hypoglycaemia was notconsistently associated with a subsequent increasein symptomatic distress or decline in diabetes-related QoL. Models using the SCL-90R showedthat only where patients had repeated severehypoglycaemia, resulting in coma or seizure, didthey tend to be at increased risk of measurablesymptomatic distress. The models using theDQOL score as the outcome did not demonstratean association of hypoglycaemia with an adversechange in QoL. The authors draw attention tosome of the factors that may account for thesefindings (e.g. lack of power to detect anassociation between hypoglycaemia and QoL asmeasured by the DQOL), but present the findingsas an indication that patients on intensive therapydid not face a deterioration in QoL, despiteincreasing demands of their diabetes care and thefrequency of hypoglycaemia (for further detail, seethe DCCT references listed).
Although severe hypoglycaemia is seen as a majorcomplication by diabetic patients, other studiesseem to support the supposition from the DCCTanalysis that hypoglycaemia may not have asignificant impact on the patients’ reported QoLassessments. Ferguson and colleagues149 in a studyto investigate the potential of insulin Lispro tolimit the frequency of severe hypoglycaemia, in acohort of patients with Type 1 diabetes who are ata high risk of severe hypoglycaemia, present someindicative findings on the relationship betweensevere hypoglycaemia and QoL. The study (anopen-label design) used the DTSQ andHypoglycaemia Fear Survey (HFS) to assess QoL,and reported that overall, considering bothtreatment groups together, the 55% of studyparticipants (18 of 33 patients) who experiencedsevere hypoglycaemia during the 48 weeks of studyfollow-up scored significantly higher on the HFSscales for worry (p = 0.049) and behaviour (p =0.015), and these patients also had significantly
Economic analysis
62
lower DTSQ scores (p = 0.040). Analysis across allpatients in the study showed an increase inexposure to severe hypoglycaemia correlated withgreater anxiety (r = 0.55, p = 0.001) and worry (r= 0.58, p = 0.001), depression (r = 0.45, p =0.010) and lower levels of energy (r = –0.52, p =0.002). However, where one treatment groupshowed a reduced incidence of severehypoglycaemia (insulin Lispro treatment), therewere no significant improvements in QoL measuresbetween the two treatment groups.
Boland and colleagues98 present findings from aprospective follow-up study in 75 adolescents andyoung adults with Type 1 diabetes, where patientschose between CSII and MDI prior to follow-up inthe study. Rates for all hypoglycaemic eventsrequiring assistance or resulting in coma werereduced by almost 50% in the CSII group (134versus 76 per 100 patient years, p < 0.05), andpatients in the CSII group found coping withdiabetes to be less difficult than patients usingMDI (p = 0.05). However, QoL assessed using theDQOLY (a version of the DCCT DQOLinstrument) showed no differences between thetwo groups. The study also reports no differencebetween groups in terms of depression (assessedusing psychosocial assessments), and self-efficacyassessments. (Note that this study was a non-randomised prospective study, and subject toselection bias, and so was not included in ourclinical review.)
Nordfelt and Jonsson140 present findings from adescriptive study to describe costs and other short-term effects of severe hypoglycaemia in childrenand adolescents with Type 1 diabetes. The study ispredominantly a cost-of-illness study, with a 12-month prospective follow-up. The study doesnot assess CSII and the study sample group(n = 129) are patients treated intensively withmultiple insulin therapy, based on active self-control, problem-based learning and psychosocialsupport. The interest in this study is the focus onhypoglycaemia and general comments on QoLissues associated with severe hypoglycaemia. Thestudy used a questionnaire containing 20 detailedquestions regarding short-term effects after eachevent of severe hypoglycaemia (during January toDecember 1988). Questionnaires were obtainedfrom 50 patients who had reported one or moreevents of severe hypoglycaemia. Nine patients(7%) reported severe hypoglycaemia withunconsciousness (U) and 45 (35%) reported severehypoglycaemia without unconsciousness butneeding assistance of another person (NU). Fourpatients had both U and NU events. Data were
collected on 111 events of severe hypoglycaemia(16 were U and 95 were NU events). With respectto QoL, patients indicated cancellations ofplanned activities in 11 events (10%),tiredness/headache in seven events (6%),schoolwork affected in five events (5%) and anger,sadness, minor wounds or bed/clothes wetting intwo events. Other effects were reported forparents/families; increased worry after nine events(8%), poor sleep/tired during the day after eightevents. As can be seen from the questionnaireresponse data, the assessment appears to havebeen primarily assessing QoL issues related toactivities of daily living and psychosocial impacts;the authors do not offer further detail on thequestionnaire administered.
Nordfelt and Jonsson140 also present brief detail ona health state valuation exercise undertaken as partof their study. The study used the EQ-5Dinstrument in a postal survey of patients with aduration of diabetes longer than 1 year. The EQ-5Dscores indicated lower global QoL for the patientswith severe hypoglycaemia in 1997 compared withthose without the severe hypoglycaemia: median0.85 versus 1.0, p = 0.0114. The authors do notoffer further detail on their findings or methods.The EQ-5D typically measures QoL on a scale of0–1, with 0 representing the worst imaginablehealth and 1 reflecting the best imaginable health,whereby respondents are asked to place a rating fortheir own health state on a visual analogue scale(VAS) from 0 to 100. Further detail is required onthis paper before findings can be generalised, assome ambiguities exist as to the comparator groupsand the methods employed to obtain values(presumably the study used the VAS section of theEQ-5D instrument).
The association between severe hypoglycaemiaand QoL differences is not apparent from theliterature. The literature does not offer areasonable estimate of health state values/utilitiesassociated with health states involving severehypoglycaemia. Some insights are available andthese will be explored in the section ‘Economicmodels for Type 1 diabetes’ (p. 66).
Fear of hypoglycaemiaBecause of the ever-present threats fromhypoglycaemia (physical symptoms, negativemood states, cognitive dysfunction and the risk ofseizures or death), many patients acknowledgesubstantial anxiety and fear concerninghypoglycaemia.150 Fear of hypoglycaemia mayitself be associated with poor metabolic control(e.g. deliberately maintaining high glucose levels
Health Technology Assessment 2004; Vol. 8: No. 43
63
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
or over-treating early symptoms).150 A surveyinstrument, the Hypoglycaemia Fear Survey(HFS),151 has been developed to assess patients’fear. The HFS is a 27-item questionnaire, with afive-point Likert scale format. The HFS was usedin a study by Ferguson and colleagues,149 discussedabove. However, the literature on the relativeeffectiveness of CSII versus MDI does not offerany detail on any differences in hypoglycaemicfear between the treatment groups.
In the appraisal of long-acting insulin analogues(insulin glargine), the NICE Appraisal Committeeaccepted that episodes of hypoglycaemia aredetrimental to QoL, partly owing to fear of suchepisodes. The Committee accepted that reductionin fear of hypoglycaemia could have a significanteffect on QoL, additional to that obtained fromreducing the QoL losses from the actual episodes.Specific QoL and utility values are not given in theappraisal consultation document for long-actinganalogues, but it appears that the cost-effectiveness estimates were significantlyinfluenced by a reduction in fear of hypoglycaemia(NICE website at www.nice.org.uk, accessed 18 October 2002).
Hypoglycaemic awarenessImpaired awareness of hypoglycaemia, defined asa reduced ability to perceive the onset ofhypoglycaemia, is common in patients with Type 1diabetes, affecting around 25% of patients,152 withprevalence increasing with duration of diabetes.149
Up to 50% of IDDM patients 15–20 years post-diagnosis report having lost their ability toperceive autonomic symptoms associated with lowblood glucose levels and thus often fail to act toprevent severe hypoglycaemia.152 Severehypoglycaemia has been reported to occur up tofive times more frequently in patients withreduced awareness of hypoglycaemia. One of thebenefits of better control through CSII may be thereturn of awareness of impending hypoglycaemia.
Clarke and colleagues152 have developed a surveyinstrument to assess patient awareness ofhypoglycaemia, and report a prospective study in78 IDDM patients. The study concluded thatIDDM subjects who believed they had reducedawareness of hypoglycaemia were generallycorrect. As part of this work, Clarke andcolleagues152 report that reduced awareness led toa greater number of moderate and severehypoglycaemic events (results were fromprospective diary records and were statisticallysignificant at p values of 0.026 and 0.0062,respectively). In support of this finding, Gold and
colleagues153 (cited by Bode and colleagues81)report a sixfold higher incidence of severehypoglycaemia in patients with hypoglycaemicunawareness compared with patients withoutimpaired awareness. The literature review on theeffectiveness of CSII versus MDI does not offerany detail on differences in hypoglycaemicawareness between the treatment groups.
CSII versus MDI: quality of lifeassessmentThe search of the literature on QoL associated withCSII identified only one study that included QoLas an outcome in a randomised trial comparingCSII with MDI.79 Tsui and colleagues79 randomlyassigned 27 patients with Type 1 diabetes to CSIIor MDI. This study reports a QoL assessment atbaseline and 9 months, as a secondary outcome,using the DQOL questionnaire. The study did notobserve any significant differences in DQOL scoresbetween the two groups (see detail on the study inAppendices 6 and 10).
There were no further findings on QoLcomparisons between CSII and MDI fromrandomised or controlled trials, identified as partof the literature review. However, we haveconsulted a number of general and cross-sectionalstudies to provide an insight to the health statusassociated with CSII treatment. We have discussedabove a number of studies that address issuesrelated to severe hypoglycaemia and QoL andoffer brief detail below of studies by Chantelauand colleagues154 and Lewis and colleagues155 thatconsidered general QoL issues in the context ofCSII. However, these studies are not randomisedor crossover designs, and the quality of the studies(often descriptive in nature) is variable; cautionshould be exercised in the interpretation andgeneralisation of any findings.
The study by Chantelau and colleagues154 presentsfindings from a prospective cohort study with a 6-month follow-up. The study uses the DCCTDQOL questionnaire to consider QoL in IDDM.The study considers two groups. Group A werepatients who had moved to a more intensivetherapy (one group is CSII). The study reports nochange on scales for social worries or diabetes-related worries, the scale for impact of diseaseshowed a reduction and the score for patientsatisfaction increased. CSII indicated fewerhypoglycaemic events. (The DQOL instrument has46 items, four subscales, scored on five-pointLikert scales.) Group B comprised a comparison ofpens versus pumps, and pen therapy scoredrelatively low on the satisfaction subscale, mean
Economic analysis
64
3.61 (SD 0.06), and higher on the impact subscale,mean 2.20 (SD 0.06). CSII significantly improved(p = 0.02) these subscale scorings only [satisfactionup to 4.02 (SD 0.06) and impact down to 2.03(SD 0.05)], with the scales for social worries anddiabetes-related worries unchanged.
Within subgroup analysis, pump users (Group Bsubgroup B1) scored significantly improved QoLwith regard to hypoglycaemia, whereas pen users(Group A subgroup A1) did not (those changingfrom pen to pump, B1, reported less frequenthypoglycaemia, as opposed to those changing fromtraditional therapy to intensive pump therapy whoreported hypoglycaemia as ‘unchanged’).
Lewis and colleagues155 present a TreatmentSatisfaction Scale for IDDM diabetes, with threesubscales. The authors offer support for the scaleand some experimental results, which indicate thatCSII has better QoL than other treatment groups.However, the authors also raise concerns over thegeneralisability of the study findings.
In a general review of QoL issues associated withdiabetes, Rubin and Peyrot142 state that it is notpossible to conclude that QoL differences (i.e.diabetes versus non-diabetes) are due to diabetesper se rather than some other characteristicsassociated with diabetes. The authors sum up bystating that better glycaemic control is associatedwith better QoL, but it is the complications ofdiabetes that are the most important disease-specific determinant of QoL. This position appearsto be the position of the WESDR and DCCTinvestigators also. Based on findings on QoL fromthese long-term studies, and from what we canlearn from the limited empirical evidence availableelsewhere, it would seem reasonable to assumethat differences between CSII and MDI in terms ofthe rates of severe hypoglycaemia may not inthemselves result in a difference in QoL, as assessedusing instruments such as DQOL and SCL-90R.
As highlighted in the patients’ perspective sectionof this report, hypoglycaemia is undoubtedly aserious concern for diabetic patients, withpotentially severe consequences, yet the overallimpact does not appear to be captured in thegeneral assessment of QoL. This may be due tothe lack of sensitivity in the instruments used, or itmay be due to the short-term nature of the eventsthemselves, or the fact that the difference betweenCSII and MDI may, on average, only reflect areduction of one event or less per patient per year.These considerations are mere speculation andshould be the subject of further research efforts.
CSII may result in a more flexible lifestyle withregard to meal schedules, work and recreation,with flexibility over the timing and content ofmeals. CSII may offer variability in timing anddose selection for insulin therapy and it may allowpatients to follow a flexible lifestyle, offering thecapability easily to correct blood glucose values byaltering infusion rates to meet swings in thesevalues.156 However, there is no evidence base todemonstrate how a reduction in the rate of severehypoglycaemia impacts on patients’ QoL.Likewise, although the fear of hypoglycaemia andthe reduced awareness of hypoglycaemia are veryreal issues for a number of patients, it is notpossible at present to quantify the impact of CSIIon the fear of hypoglycaemia or the effect of CSIIon the ability of patients to perceive the onset ofhypoglycaemia (let alone the QoL issuesassociated with such impacts).
Health state values/utilities for CSIIThe literature identified did not contain anyinsight as to the values, or health state utilities,that patients ascribe to the use of CSII.Furthermore, we found more generally that theliterature on health state values associated withdiabetes was also sparse, although this was subjectto ad hoc searches only and by a systematic reviewof the literature.
Wu and colleagues157 present analyses on healthstate values for diabetes derived via a mappingprocess, from SF-36 responses to the valuesavailable from the QWB Scale. The findings arebased on analysis from a sample of 89 (non-diabetic) respondents completing the SF-36. Thepaper is a methods paper and can be classed asexperimental; however, the presentation ofestimates of QWB scores associated with a movefrom ‘general population health state values’ to acondition in which patients are ‘Type 1 diabetics,with no complications’, and from the ‘nocomplications’ diabetic state to a state involving‘diabetic retinopathy’ may be helpful for thedevelopment of illustrative analysis within thisreview. Table 33 presents outline findings from thestudy by Wu and colleagues. Caution must beexercised when considering the data presented inthis study.
Data from Wu and colleagues157 indicate thathealth state values associated with different statesshow only small differences in valuations (e.g. fora move from ‘no complications’ to ‘retinopathystate’ we see a change of 0.03, 0.04 and 0.02 forthe groups in Table 33). The clinical significance ofthese differences is dubious, and retinopathy
Health Technology Assessment 2004; Vol. 8: No. 43
65
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
would only affect QoL if it impaired vision, or if itled to anxiety though the diagnosis per se, or fromdisutility from treatment given.
With regard to diabetic retinopathy, Brown andcolleagues158 assessed the utility values associatedwith varying degrees of visual loss due to diabeticretinopathy. The study presents utility values,elicited using standard gamble (SG) and timetrade-off (TTO) techniques, across five subgroupswith varying degrees of visual loss, ranging from0.85 to 0.59 for TTO and from 0.70 to 0.90 for SGscores. Overall, in the sample of 95 respondentsthe TTO values were 0.77 and the SG scores were0.88 (with visual acuity ranging from 20/20 visionto hand motion visual acuity in the best eye).
Kiberd and Jindal,159 in a study on screening toprevent renal failure in insulin-dependent diabeticpatients, present an estimate of the health stateutility for patients with diabetes; they report avalue of 0.838, commenting that traditionallyutilities vary between 1.0 (perfect health) and 0(death). The authors, finding no publishedliterature to inform on health state values,determined these values using a TTO format in asample of 17 healthcare workers not associatedwith their study (this sample consisted ofnephrologists, clinical house staff, nurses and onesocial worker). The sample of healthcare workersestimated values for six health states, one of whichwas ‘insulin-dependent diabetes alone’. Theauthors do not report any further detail on thehealth state valuation exercise.
The above studies have no direct relevance to thecomparison of CSII and MDI, but they do offersome idea of the magnitude of the health statevaluation differences between diabetes-relatedhealth states, and may offer an opportunity toconsider how the difference between CSII andMDI might relate to the differences associatedwith (i) a non-diabetic state versus a diabetic stateor (ii) a diabetic state with no complications versus
a diabetic state with retinopathy complications.Once again, in raising these matters forconsideration we stress that caution must beexercised in generalising from these essentiallyexperimental studies or using the information foranything other than illustrative purposes.
Economic models for Type 1diabetesReview of economic models for Type 1diabetes mellitus (IDDM)Given the limited data available to inform on thecost-effectiveness of CSII versus MDI, a generaltopic search of the diabetes literature wasundertaken to identify literature on the model-based economic assessments within Type 1diabetes (IDDM). Only a limited number ofmodel-based approaches have been identified thatassess economic outcomes for Type 1 diabetes.Table 34 provides summary detail on themodelling approaches associated with Type 1diabetes. Models are described in outline, with aparticular focus on diabetic retinopathy andhypoglycaemia, where they form part of the modelstructure. The models are described to offerbackground to the assessment of CSII but do notdirectly assist with the assessment of CSII versusMDI.
DCCT Research GroupThe DCCT has been discussed above. The DCCTmodel129 describes and compares the lifetimebenefits and costs of conventional and intensivetherapy as implemented in the DCCT. The modelis a Monte Carlo simulation model, used topredict the incidence of microvascular andneurological complications in a hypotheticalsample of 10,000 persons with IDDM. It randomlyselects from the hypothetical population (either aprimary prevention cohort, or secondaryprevention) and assigns characteristics (e.g. age,disease characteristics). It then uses 12 health
Economic analysis
66
TABLE 33 Age- and health-specific QWB scores. (Copyright © 1998 American Diabetes Association. From Diabetes Care 1998; 21:275–31.157 Reprinted with permission from The American Diabetes Association.)
Age (years) General populationa Type 1 diabetic, Type 1 diabetic, Otherb
no complications with retinopathy only
<45 0.82 0.73 ± 0.07 0.76 ± 0.05 0.70 ± 0.0845–64 0.75 0.68 ± 0.09 0.72 ± 0.09 0.66 ± 0.07≥ 65 0.70 0.64 ± 0.08 0.62 ± 0.07 0.55 ± 0.05
a Data on general population are from previous studies; see Wu et al.157 for detail.b Type 1 diabetic individuals with diabetic neuropathy or nephropathy alone, or with other complications.
states to capture disease characteristics, groupedaccording to the three major complicationsstudied in the DCCT (retinopathy, neuropathy,nephropathy). The model simulates the course ofthe patient’s disease over their expected lifetime.The model uses 1-year cycles and at each cycle anindividual is in one of five retinopathy healthstates, one of four nephropathy health states andone of three neuropathy health states. Theprobability that a patient will advance to a moresevere stage of disease in a given year depends onthe patient’s current state of health, treatmentregime (i.e. intensive versus conventional insulintherapy) and treatment duration. The modelcycles through time at a patient level, until thepatient exits the model (due to death) and thenthe next patient is selected from the hypotheticalsample. This process is repeated in the DCCTanalysis for a sample of 10,000 individuals. At theend of the modelling process (the simulation), thetime spent in each of the treatments and healthstates and the time spent alive are calculated, costsare assigned and mean statistics are calculated bytreatment group (conventional versus intensive).The DCCT model does not considerhypoglycaemic events.
The DCCT model uses empirical data on diseaseprogression, over 9 years, from the DCCT and aseries of statistical models (Weibull models) topredict the probability of patients advancing tobackground diabetic retinopathy (BDR) andproliferative diabetic retinopathy (PDR) [e.g.Weibull model, � × � × t(� – 1), where � and � arestatistical parameters determined by the study andt is the parameter for duration of treatment;different � and � parameters were determined toreflect conventional and intensive treatmentprobabilities of progression of disease]. Thesemethods are not transferable to the assessment ofCSII versus MDI. Table 35 offers outline detail onthe DCCT model structure.
Palmer and colleaguesThe diabetes disease model developed by Palmerand colleagues135 considers the cost-effectiveness
of a range of intensive interventions for Type 1diabetes (IDDM) compared with conventionaltherapy, to consider optimal lifetime treatmentpatterns. A variant of this model has been used inan earlier NICE submission on pioglitazone inType 2 diabetes.171 (Chilcott and colleagues,2001). The Type 1 model of Palmer andcolleagues is a microsimulation model, simulatingthe experiences of individual patients (themodelling process is similar to the processdescribed above for the DCCT model). The datafor the model are generally drawn from the DCCTand WESDR studies. The authors cite an earlyWESDR study144 and the main DCCT study19 asevidence that the rate of disease progression fromno retinopathy to BDR is dependent on durationof diabetes and the HbA1c value;135 yet the modeluses DCCT methods surrounding duration ofdiabetes to predict transit probabilities for BDR.19
Palmer and colleagues’ model comprises a seriesof Markov submodels, representing thedevelopment and consequences of renal disease,retinopathy, amputation, myocardial infarction,stroke, major hypoglycaemic events and DKA. Thedata used in the model on hypoglycaemic eventsare data (event rates) from the DCCT, reflecting agreater risk of severe hypoglycaemic events for theintensive treatment group versus conventionaltherapy.
The submodel for hypoglycaemia (Figure 12) is asimple transition from an entry state involving nomajor hypoglycaemic event to a health statewhereby patients experience an event requiringassistance (severe hypoglycaemia) and then transitback to either the health state of no hypoglycaemiaor hypoglycaemic death (the model assumes a verysmall probability of death as a consequence ofhypoglycaemia, a death rate of 0.0001).
The submodel for diabetic retinopathy is describedin Figure 13.
For retinopathy, the pattern of disease progressionis from no retinopathy to BDR, to PDR and on to
Health Technology Assessment 2004; Vol. 8: No. 43
67
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
TABLE 34 Approaches to model the cost-effectiveness of Type 1 diabetes
Study Study design Approach Intervention
DCCT, 1996129 Modelling Cost-effectiveness Conventional versus intensive therapy
Palmer, 2000135 Modelling Cost-effectiveness Conventional versus intensive therapy(various treatment options)
Tomar, 1998160 Modelling (based on Cost-effectiveness Conventional versus Intensive therapyDCCT model above) (plus costing study)
blindness (Figure 13). As the simulation progresseseach year (1-year model cycles), patients maycontinue in their present health states or transit tothe next stage of disease progression or a non-specific mortality state. Transit probabilities arebased on DCCT data and are dependent onduration of diabetes. Therefore, the methodologyis not transferable to an evaluation of CSII versusMDI. The retinopathy submodel does not includemacular oedema. The retinopathy submodelmakes assumptions surrounding screening andtreatment for diabetic retinopathy.
The model examines the course of thedevelopment of diabetes complications under a set
of intervention scenarios and examines thetreatment effects in terms of a time series ofclinical events and outcomes (e.g. incidence ofcomplications, mortality). It produces results for atypical cohort of 19-year-old diabetic patients, withno baseline complications (to reflect a typicalcohort of Swiss male patients). It does not have aseparate submodel for mortality, as mortality isincluded within each of the complication specificsubmodels. However, it is not clear how themortality rates have been determined. Table 35offers outline detail on the model structure.
Tomar and colleaguesThe model by Tomar and colleagues160 is notdescribed in this report as it is based on theapproach documented by the DCCT ResearchGroup (as above), and does not offer additionaldata to inform on the modelling of diabetes forCSII versus MDI. See the cited reference forfurther detail.
Suitability of economic models forevaluation of CSII versus MDIThe modelling approaches discussed above do notoffer an opportunity to transfer methods to theassessment of CSII versus MDI. The model-basedapproaches described for Type 1 diabetes (andmuch of the Type 2 diabetes literature) areconfigured to assess the impact of intensivetherapy compared with conventional therapy,whereas the assessment of CSII versus MDI is acomparison of different treatment options for thedelivery of intensive therapy.
Economic analysis
68
Nohypoglycaemic
event
Severe hypoglycaemiarequiring medical
assistance
Death fromhypoglycaemic
event
FIGURE 12 Hypoglycaemia submodel from Palmer andcolleagues135
No retinopathy
Backgroundretinopathy
Non-specificmortality
Blind
Proliferativeretinopathy
FIGURE 13 Retinopathy submodel from Palmer and colleagues135
Illustrative analysis of the relationshipbetween HbA1c reductions and theincidence of diabetic retinopathyGiven the lack of models with which to comparedirectly the cost-effectiveness of CSII with MDI,one option is to use the observed mean reductionin HbA1c in combination with the DCCT modelfor a somewhat speculative estimate of reductionin future retinopathy.
The DCCT Research Group demonstrated that aregimen of intensive therapy aimed at maintainingnear normal blood glucose values markedlyreduces the risk of development and progressionof retinopathy and other complications of IDDMwhen compared with a conventional treatmentregimen.19 The DCCT Research Group alsopresent a further epidemiological assessment ofthe association between levels of glycaemicexposure (HbA1c) before and during the DCCTwith risk of retinopathy progression within eachtreatment group.143 Total glycaemic exposure(HbA1c and the duration of exposure) was thedominant factor associated with the risk ofretinopathy progression.143 When examinedsimultaneously within each treatment group, eachof the components of pre-DCCT HbA1c, pre-DCCT duration of IDDM, mean HbA1c during the
study, time in the study and their interaction, wassignificantly associated with risk of retinopathyprogression.143 However, the DCCT data cannotprovide a definitive assessment of the causalrelationship between specific levels of glycaemicexposure and the risk of complications. The studywas designed to assess whether an intensiveregimen aimed at lower levels of glycaemia wouldyield the desired effects (i.e. reducedcomplications), recognising that any effectsobserved could theoretically be attributable to anyof the effects of such therapy.143 However, theDCCT Group do state that the epidemiologicalassessment of data conclusively demonstrates thatthe updated mean HbA1c (i.e. followingintervention) during the trial and the years offollow-up in the study are the most importantpredictors of the risk of complications in theconventional treatment group, and that theupdated mean HbA1c was the dominant predictorof risk of complications in the intensive treatmentgroup. Although glycaemic exposure was thedominant predictor, the updated mean HbA1c didnot completely explain the risk of progression.
The DCCT Research Group143 investigatedwhether there was a threshold for glycaemia belowwhich there was no further reduction in risk of
Health Technology Assessment 2004; Vol. 8: No. 43
69
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
TABLE 35 Models for IDDM and model characteristics
Modelling studies
Detail DCCT Research Group129 Palmer and colleagues135
Study design Modelling Modelling
Economic outcomes Cost-effectiveness Cost-effectiveness
Intervention Intensive insulin therapy Intensive therapy combinations
Type of modelling Microsimulation (Monte Carlo, Microsimulation (Monte Carlo, patient level, 1-year cycles) patient level, 1-year cycles)
Markov model Markov model
Decision analysis Yes Yes
Cohort/age Two cohorts aged 13–39 years Swiss cohort of 19-year-old diabetics (primary/secondary prevention) patients
Source of cohort information WESDR
Complications:Retinopathy � �Neuropathy � �Nephropathy � �Heart Disease – �Stroke – �Hypoglycaemia – �Ketoacidosis – �Data inputs for retinopathy WESDR DCCT/WESDRData inputs for hypo – DCCT
complications, finding that over the range ofvalues for intensive treatment, there was a gradualdecline in risk with each additional reduction inHbA1c (i.e. no threshold was seen below which therisk of retinopathy was eliminated entirely). Theauthors present a number of statistical models(regression models) used to predict the risk ofretinopathy progression. In Figure 14 we apply oneof these models in an illustrative analysis to assessthe predicted risk of retinopathy progressionamongst patient groups subject to CSII or MDI.
Coefficients for this model are presented by theDCCT Group as shown in Table 36.
The exploratory model (see Figure 15), uses a timehorizon of 10 years (10 years has been used as theanalysis does not accommodate mortalityconcerns) to simulate patient level experiences forCSII and MDI therapies against differing HbA1c
profiles, to reflect the efficacy parametersexpected from CSII versus MDI. A simple Excelspreadsheet model has been used, using CrystalBall software (Decisioneering®) to produce asimulation.
We have applied the above statistical model[equation (1)] to predict the probability of onset ofsustained progression of diabetic retinopathy in a
population defined by baseline HbA1c andduration of IDDM at the start of therapy. Thecohort is assumed to be a primary preventioncohort, free from diabetic retinopathy. Cohort datahave been gathered from studies within theclinical review described above (see Figure 15 fordetail). The efficacy data from the clinical reviewhave been used to reflect a difference in baselineHbA1c and follow-up HbA1c. To simulate theprogression of disease over time, the model usesthe transit probabilities determined through use ofthe statistical model and appropriate modelcoefficients, together with a series of randomnumbers, with the onset of disease based on thebalance of probabilities within the patient cohort.The onset of sustained progression of disease is asdefined in the DCCT study.
A simple simulation has been run over the 10-yeartime frame, to predict the experiences of a patientcohort with CSII (reflecting a reduced level ofHbA1c) and a patient cohort on MDI (i.e. noreduction in HbA1c), using the same series ofrandom numbers to cycle both patient groupsthrough the model/simulation. A simulation hasbeen carried out using a group of 10,000 patients.Results generated from specific input parameters(i.e. HbA1c and duration of IDDM) are presentedin Table 37, for illustrative purposes only [this
Economic analysis
70
Absolute risk at time (t) = f (HbA1c, log of duration of IDDM on entry to the study, time)
�(t) = exp{� + �1 (H) + �2 ln(D) + �3 (t) + �4 ln(A) + [�5(t) × ln(A)]} (1)
where �(t) is the absolute risk at time t, � is an intercept term, �1 is the screen HbA1c coefficient, H is the screeningHbA1c, �2 is the duration coefficient, D is the duration of IDDM (in years) on entry, �3 is the time coefficient, t is thestudy time and �4 and �5 are interaction coefficients. Hence,
� (risk of onset of sustained retinopathy progression at t) = exp{� + �1 (screen HbA1c) + �2 (log duration of IDDMon entry) + �3 (time) + �4 (log mean updated HbA1c) + [(�5 × time) × (log mean HbA1c)]} (2)
FIGURE 14 DCCT regression model used for illustrative analysis
TABLE 36 From DCCT143 Table 9: regression models of sustained retinopathy progression as a function of total glycaemic exposure
Intensive – treatment
Total glycaemic exposure, Poisson model Estimate 95% CI p
Intercept –10.871 –16.964 to –4.779 <0.001Screening HbA1c (%) 0.215 0.061 to 0.37 –Log (duration) (years) 1.055 0.656 to 1.454 –Pre-DCCT exposure (2 df) – – <0.001Time (years) – 1.710 –3.120 to –0.300 –Log (mean HbA1c) 1.541 –1.46 to 4.541 –Log (mean HbA1c) × time 0.809 0.136 to 1.483 <0.019
exploratory analysis is based on an assumption ofprimary prevention cohorts (diabetic retinopathy),assuming that patient cohorts are free fromdisease].
Cost-effectiveness of CSII versusMDI We have discussed above the difficulties associatedwith modelling cost-effectiveness via a diseaseprogression model using glycated haemoglobin(HbA1c), the main effectiveness outcome from thecomparison of CSII and MDI. From ourexploration of the literature and availablemodelling technologies, it is not possible toproduce a meaningful incremental cost-effectiveness ratio. We have produced a costanalysis to determine the short-term and medium-term costs associated with CSII. However, we havenot been able to identify comprehensive estimatesof meaningful health outcomes from the clinicalevidence available. The link between HbA1c andlonger term complications (e.g. eye and kidneydisease) would have to be assumed. There is anabsence of data linking the use of pumps with anylonger term complication or mortality data. Toour knowledge there are no long-term trials ofCSII versus MDI planned for the future.
Below we make some assumptions surrounding thebenefits associated with CSII. However, based onthe lack of an evidence base to support theseassumptions, the examples are purely illustrative.
Outcomes associated with CSIILong-term complicationsWe have discussed above the general belief that areduction in HbA1c is a good thing, with areduction in HbA1c expected to contribute to areduction in the incidence of complications suchas diabetic retinopathy (DCCT,19 UKPDS,131
WESDR132). However, in previous studies assessingthe cost-effectiveness of interventions for Type 1diabetes (DCCT,129 Palmer and colleagues135),hazard rates and statistical parameters have beenused that have been determined from particularstudy and intervention groups, in order toestimate future rates for complications and thesubsequent reduction in risk associated withdiabetic interventions (e.g. intensive therapyversus conventional therapy). Pickup andcolleagues40 in a recent meta-analysis of CSII havealluded to a potential reduction in the onset ofsustained retinopathy progression and we haveundertaken some exploratory analysis to assesswhether their claims are supported by the datafrom the DCCT. Using a statistical model to assessthe absolute risk of sustained onset of retinopathy
Health Technology Assessment 2004; Vol. 8: No. 43
71
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Assign patient-levelcharacteristicsfrom cohort data
Simulation (10,000 patients)
No DR
Patients go through 10 × 1-year cycles with risk of onset of sustained progression of DR
NoYesOnset of DR(or cycle = 10)
Patients either exit the model free of DR or subject to onset of DR
Ass
ign
next
pa
tient
CohortDescription– HbA1c (entry)– Duration of DM (on entry)
FIGURE 15 Structure of exploratory model to assess risk of onset of sustained progression of diabetic retinopathy (DR)
Economic analysis
72 TA
BLE
37
Illus
trat
ive e
xam
ples
of t
he re
duct
ion
in ri
sk o
f ons
et o
f pro
gres
sion
of d
iabe
tic re
tinop
athy
in a
coh
ort
of p
atie
nts
free
from
retin
opat
hy
Coh
ort
char
acte
rist
ics
Sour
ce o
f cha
ract
eris
tics
MD
I –
onse
t of
C
SII
– on
set
of
Ris
k re
duct
ion
Bas
elin
e H
bA1c
: D
urat
ion
of
Effic
acy
para
met
erdi
seas
e: 1
0-ye
ar d
isea
se: 1
0-ye
ar10
-yea
r (%
)A
vera
ge a
nnua
l m
ean
(SD
) di
abet
es: m
ean
(red
ucti
on in
pr
obab
ility
(%
)pr
obab
ility
(%
)re
duct
ion
(%)
[ass
umpt
ion:
(S
D; r
ange
) H
bA1c
)no
rmal
[a
ssum
ptio
n:
[ass
umpt
ion:
dist
ribu
tion
]lo
g-no
rmal
no
rmal
di
stri
buti
on]
dist
ribu
tion
]
8.39
(0.8
7)20
(11.
2; 4
–42)
–0.5
2aH
anna
ire-B
rout
ine
et a
l.8052
.23
42.2
99.
940.
9910
.4 (1
.85)
9.2
(5.5
5)–0
.52a
Haa
kens
et
al.83
70.9
464
.41
6.53
0.65
7.73
(0.6
)17
(10;
8–3
7)–0
.52a
Tsui
et
al.79
– C
SII c
ohor
t ch
arac
teris
tics
30.8
223
.15
7.67
0.77
8.16
(0.7
)15
(9; 4
–28)
–0.5
2aTs
ui e
t al
.79–
MD
I coh
ort
char
acte
ristic
s37
.96
29.0
18.
950.
90
Som
e va
lidat
ion
of t
he m
odel
use
d ha
s be
en a
ttem
pted
thr
ough
est
imat
ion
of t
he r
isk fo
r th
e D
CC
T p
rimar
y pr
even
tion
coho
rt, b
ased
on
data
ava
ilabl
e in
DC
CT:
19,1
29
9.1
(1.6
)2.
6 (1
.4)
–1.9
%D
CC
T –
inte
nsiv
e tr
eatm
ent,
prim
ary
Mod
el p
redi
cts
15%
risk
of o
nset
(ove
r 10
year
s) b
ased
on
coho
rt
(ass
ume
SD
prev
entio
n co
hort
char
acte
ristic
s, a
nd m
ean
effic
acy
data
. The
dat
a pr
esen
ted
in t
he
of 0
.19)
DC
CT
indi
cate
d a
rate
of 1
.2 p
er 1
00 p
atie
nt y
ears
, the
refo
re t
hem
odel
is n
ot fa
r fr
om t
his
estim
ate
(i.e.
1.5
% fr
om t
he m
odel
ver
sus
1.2%
from
the
DC
CT
dat
a pr
esen
ted
by t
he D
CC
T G
roup
19)
Oth
er s
tudi
es, e
.g. N
osad
ini e
t al
.,85Bo
de e
t al
.,81Sc
hiffr
in e
t al
.,97Z
iegl
er e
t al
.,84da
ta a
vaila
ble
do n
ot a
llow
inve
stig
atio
n.a
Effic
acy
data
from
the
met
a-an
alys
is pr
esen
ted
in C
hapt
er 3
, cov
erin
g tr
ial d
ata
over
10
wee
ks t
o 4
mon
ths,
from
incl
uded
RC
Ts a
nd C
CTs
; mea
n re
duct
ion
0.52
% (S
D o
f 0.2
2as
sum
ed t
o re
flect
the
rep
orte
d 95
% C
Is o
f –1.
19 t
o 0.
14).
progression, a Poisson model detailed by theDCCT Research Group, we would concur with thesuggestion of Pickup and colleagues that a 0.5%reduction in HbA1c would result in a reduction inthe region of 0.5 cases of sustained retinopathyprogression per 100 patient years.
The analysis we have undertaken to relate theeffect of HbA1c changes to the incidence of long-term complications is illustrative. The statisticalmodel used does not differentiate between thenature of the outcome expected, sustainedretinopathy progression. BDR, which wouldcomprise most, if not all, of this forecast reductionin the risk of diabetic retinopathy, is very differentfrom PDR in terms of both its economicimplications and the implications for the diabeticpatient. The onset of BDR does not involveadditional NHS resources as there is no treatmentother than insulin therapy to manage glycaemiccontrol. All diabetic patients should be subject toeye screening to detect onset and progression ofdiabetic retinopathy. The probability of onset ofBDR is relatively high compared to the progressionfrom BDR to PDR, hence our concern over theability of the statistical model available to modeldisease progression. Our exploratory analysis ofthe methods presented by the DCCT ResearchGroup143 involved a simulation of patient groupsdefined by levels of HbA1c and duration of diabetesover a 10-year period to detect onset of BDR froma population assumed to be free from BDR. Fromthe clinical studies identified to inform on theeffectiveness of CSII versus MDI we used patientcharacteristics, together with the mean effectivenesson HbA1c identified via the meta-analysis reportedabove (we assume a reduction in HbA1c of 0.52%).Findings indicate that a reduction in the risk ofonset of BDR is somewhere between 0.5% and 1%per patient per year. The effect of such areduction in the onset of sustained retinopathyprogression requires consideration in the contextof the overall incidence of disease and the impactof disease. The incidence of BDR is reported to bedependent on the duration of diabetes(WESDR,132 DCCT19), and an estimate of theannual incidence of BDR from ‘no retinopathy’would be in the region of 22% per year, based ondata from WESDR, reported in their study on the4-year incidence of diabetic retinopathy (datashowed an incidence rate of 59% over 4 years).The DCCT Research Group19 report in theirprimary prevention cohort a rate of 1.2 events per100 patient years for onset of BDR. However, thisgroup consisted of recently diagnosed patientswith a mean duration of diabetes of 2.6 years(±1.4), hence a low-risk group.
In our opinion, it would seem reasonable toassume a small reduction in the onset of sustainedretinopathy progression, predominantly from nodisease to BDR; however, the impact of any sucheffect would be small on any cost-effectivenessfindings for a population or patient analysis.Further effects on other long-term complicationssuch as nephropathy (kidney disease) may berelevant, although we are not able to identify anyliterature concerning the consequences for suchevents in the context of the size of difference inHbA1c related to the comparison of CSII and MDI.
It might also be worth noting that the move tosystematic screening for early retinopathy,followed by laser photocoagulation, will diminishthe incidence of visual loss. Hence the benefits oftighter control will, at least for retinopathy, be lessthan in past decades.
Adverse events – severe hypoglycaemiaThe impact of severe hypoglycaemic events onQoL is not known. Indications from the limitedliterature indicate that overall QoL, as measuredby instruments such as the DQOL, is not affectedby events. Events are very short term, with theacute impact of a severe hypoglycaemic eventlasting for 24 hours, at most, in most cases.
In order to assess the impact of severehypoglycaemic events on the health state utilityexperienced by the diabetic patient, we canconsider illustrative examples in a number of ways:
1. We could take the data reported by Nordfeltand Jonsson140 showing a reduction of 0.15 (ona health metric scale of 0–1), between a statecharacterised by severe hypoglycaemia and onewhich is not. As mentioned earlier, these dataare not supported by detail on methods withinthe paper, so caution must be exercised.However, it could be assumed that the healthgain associated with preventing one eventcould be characterised by dividing 0.15 (thehealth gain) by 365 days to ascertain theassociated impact for the 24-hour acute period.This obviously results in a very small differencein overall health utility, and even assuming thatthe effect of the severe hypoglycaemic eventcovered a period of 4 days, the health utilitygained from preventing an event would be0.00164, less than a 0.2% difference.
2. We could consider the health state valuesoffered as part of the EQ-5D tariff161 andassume that the effect of a severehypoglycaemic event would be reflected in a movement in the EQ-5D health
Health Technology Assessment 2004; Vol. 8: No. 43
73
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
descriptive system against the dimensions ofanxiety/depression and/or usual activities, asintuitively both dimensions may be affected byan episode of severe hypoglycaemia.Considering that an event may cause amovement on the EQ-5D algorithm from amoderate to a severe level (level 2 to level 3),or from a level associated with no problems oranxiety to a level associated with a moderatelevel of anxiety/depression and/or someproblems with usual activities (level 1 to level2). In such a way we can estimate the healthgain associated with preventing one severehypoglycaemic event. The health gainidentified in this hypothetical manner wouldstill be very small, assuming that an acute affectfrom an event covers a 24-hour period.Assuming that a patient moves from an EQ-5Dhealth state described as having no anxiety anddepression and no problems with usualactivities to a health state characterised byextreme anxiety/depression and unable to dousual activities (all other dimensions remainingconstant), the EQ-5D tariff values would reflecta difference in health utility/value of 0.294quality-adjusted life-years (QALYs). Applyingsuch a change to a short-term experience suchas a severe hypoglycaemic event lasting24 hours, we have 0.294/365, which is 0.0008QALYs per event.
3. Even if we were crudely to assume that apatient experiencing a severe hypoglycaemicevent were to move, on the health metric scaleof 0–1 (where 1 represents full health), fromfull health (score of 1) to worst imaginablehealth (score of 0), it would still only reflect ahealth gain (QALY value) of 0.00274 per event(assuming the event were to last for 24 hours).
The above considerations are purely speculativeand have no basis in the published literature, andcombining these assumptions with available costestimates (in effect assuming that the reduction in
hypoglycaemic events was the only benefit of CSII)would result in very extreme, and potentiallyunrealistic, cost-effectiveness findings(~£400,000–500,000 per QALY).
Other benefits of CSIIChapter 4 indicated that CSII is reported by somepatients to offer lifestyle benefits andopportunities to manage their diabetes in a moreflexible way. In this review we have been unable toquantify such benefits for the purpose of economicanalysis. This does not mean that such benefits arenot of importance for the patients themselves, butjust that the research necessary to quantify anyrelative utility values has not been undertaken.
Estimating the cost-effectiveness ofCSII versus MDICost estimates have been detailed above (seesummary in Table 38). In order to present cost-effectiveness estimates, data on meaningful healthoutcomes are required. We have been unable toidentify health outcomes that can be quantified forthe purposes of cost-effectiveness analysis. Basedon the cost estimates calculated as part of thisreview, and the speculation as to the health gainassociated with the prevention of severehypoglycaemia (for example, a reduction of oneevent per year), with estimated QALY values perevent ranging from 0.00164 to 0.00274, any cost-effectiveness estimates would be circa~400,000–500,000 per QALY. A summary ofbenefits/outcomes from CSII versus MDI ispresented in Table 39.
Cost-effectivenessCost per severe hypoglycaemic event avoidedThe evidence on the rate of severe hypoglycaemicevents, CSII versus MDI, is conflicting. Using datafrom Bode and colleagues,81 which cover thereduction in events associated with CSII in apatient group that has a relatively high risk ofevents (i.e. a relatively high estimate of
Economic analysis
74
TABLE 38 Estimated additional costs associated with CSII versus MDI: mean patient-level costs for 1, 4 and 8 years (figures inparentheses where future costs have been discounted at 6%)
Total net cost for CSII Disetronic D-Tron (£) Disetronic H-Tron (£) MiniMed 508 (£)
Year 1 3,878 3,571 3,602
Assuming 4-year pump life:Years 1–4 (discounted) 7,081 (6,722) 6,569 (6,242) 6,058 (5,790)Years 1–8 (discounted) 13,941 (11,871) 12,917 (11,011) 11,894 (10,201)
Assuming 8-year pump life:years 1–8 (discounted) 12,178 (10,429) 11,272 (9,663) 10,096 (8,728)
effectiveness), there is an assumed reduction of1.16 severe hypoglycaemic events per patient peryear (12-month data showed 22 events for CSIIcompared with 138 events for MDI). Applying thecost data in Table 38 to the data on severehypoglycaemia from Bode and colleagues offersan estimate of the cost per severe hypoglycaemicevent avoided. However, this is an intermediateoutcome and does not describe the health impactwith respect to a severe hypoglycaemic event.
Estimates are also presented based on an estimateof effectiveness that is expected to be at the highend of the expected benefits from CSII therapy.
Cost per severe hypoglycaemic event avoided ispresented in Table 40.
Further estimates on the cost-effectiveness of CSIIversus MDI are not possible given the dataavailable.
Health Technology Assessment 2004; Vol. 8: No. 43
75
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
TABLE 39 Summary of benefits/outcomes from CSII versus MDI
Benefit/outcome Quantifiable Related to cost-effectiveness
Reduction in HbA1c Yes Not able to link directly to diabetic complications
Reduction in severe hypoglycaemia Yes Not able to link to cost-effectiveness in a meaningful manner (i.e.only able to consider cost per event avoided). Not able toestablish QoL impact associated with the reduction in severehypoglycaemic events
Long-term complications No Not able to link HbA1c directly to diabetic complications (e.g. diabetic retinopathy) (illustrative analysis provided). Small reduction in incidence of
microvascular complications expected
QoL benefits No Insufficient data to make a judgement on overall QoL differencesbetween CSII and MDI
Lifestyle benefits/flexibility No No research available with quantifiable utility effects
TABLE 40 Cost per severe hypoglycaemic event avoided
Time horizon for analysis Cost per severe hypoglycaemic event avoided (£)a
Year 1 3078–32644-year time horizon (years 1–4)b 1305–1526 [1275–1481]c
8-year time horizon (years 1–8)b 1281–1502 [1157–1346]c
a Range reflects the differences between the three insulin pumps available.b Assuming an insulin pump life of 4 years.c Where future costs have been discounted at 6% and future outcomes discounted at 1.5%.
The cost to the NHS would depend mainly onthe proportion, and therefore the number, of
patients who used pumps. The largest cost, asshown in Chapter 5, would be those of theconsumables such as infusion sets, with the pumpbeing second. These two together cost fromaround £1075 per annum assuming an 8-year lifeof pump and the cheapest of the three, to £1423per annum assuming a 4-year life and the mostexpensive model.
Numbers would be limited. CSII is not just apump, but a package of self-care involvingfrequent self-monitoring of blood glucose andwillingness to accept further training andresponsibility. The presence of the device for24 hours will deter some people. Conversely, theyounger generation of children familiar withpocket technologies such as games, phones andhand-held computers might have a faster uptake.
Various guides to identifying patients most able tobenefit have been produced. Those from theDiabetes Federation of Ireland note that thedecision is often made by patients:
“Patient selection
� Patients tend to self-select for pumps. They areexclusively patients with Type 1 diabetes who arehighly motivated and have developed an expertisein their own condition.
� Most will have been using MDI but have notachieved their desired level of control.
� Many have early complications such as retinopathy,nephropathy or large vessel disease, whichcontribute to the patient’s personal motivation totake control of the condition. This is a motivator,not a criterion, as ideally intensive insulin therapyshould aim to prevent complications.
� Patients make an important trade-off in acceptingthe long-term attachment to the pump in exchange for better control and a lower risk ofcomplications.”
T. O’Sullivan, Diabetes Federation of Ireland,September 2001
Similar thoughts were expressed by Gerich in1985,162 who added that patients must be capableof accurate self-monitoring of blood glucose levelsand must be willing to do this several times a day.
It is unlikely that more than a small proportion ofpeople with Type 1 diabetes would become pumpusers, but this proportion is not known withaccuracy. Experience from the first few years ofthe West Kent service, provided by the Guy’sHospital trust, suggests that perhaps only 2% willmove to CSII (Pickup J: personal communication).It is likely that CSII would be used mainly by thosewith particular problems with control ofhypoglycaemia, rather than high HbA1c. It seemssafe to assume that patients with Type 2 diabeteswill not be treated with CSII, except as part ofresearch projects on reversing insulin resistance.Figure 16 shows the system used in the serviceprovided to West Kent Health Authority (Pickup J:personal communication) and Table 41 gives thecriteria used to assess suitability.
There are about 330,000 people with Type 1diabetes in England and Wales. The annual costsfor varying percentages are shown in Table 42.However these figures may be too conservative. Itis reported that in Sweden in 1999, 7.5% ofchildren and teenagers were on CSII, but that thishas now increased to 12%.163
Costs other than disposables and pumps areminor. There will be modest savings in insulindosage and from reduced hypoglycaemic events.In the longer term, there will probably be savingsfrom reduced long-term complications, althoughthese are difficult to quantify (see Chapters 5 and7). There will also be education costs, for bothpatients and staff. We have included these in ourcostings.
One issue would be the number of centres thatprovide a CSII service. Should all large diabetesclinics provide a service, or should there be alimited number, with perhaps one pumps clinicper 500,000 population, except in less denselypopulated areas where that would impose travelcosts on patients? One might expect that serviceswould be better if they looked after a minimumnumber of pump users, but there is no hardevidence for that, or for what the minimumnumber should be. The converse argument is thatall hospitals providing a diabetes clinic shouldoffer pump therapy, since it is just another way ofgiving insulin. Furthermore, if pump services were
Health Technology Assessment 2004; Vol. 8: No. 43
77
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Chapter 6
Implementation
Implementation
78
Hospital consultantGPAssessment atPump Clinic byConsultant(15 pts/yr/HA)
Poor control on MIT(two-thirds)– frequent hypo– dawn phenomenon
Pump nurse/dietitianOptimisation of controlon MIT
Unsuitable for CSII (one-third)– non-compliant– psychological problems– good control on MIT
Further treatment(e.g. psychotherapy)General Diabetic Clinic
Trial of CSII
Improved controlcontinued CSII(~10 pts/yr/HA)
FIGURE 16 Strategy for management of potential insulin pump patients
TABLE 41 Selection criteria for a trial of CSII
Type 1 diabetic patients who have failed to achieve good glycaemic control after a 3-month trial of intensive insulin injectiontherapy, including re-education in injection technique, dietary advice and blood glucose self-monitoring), because of:
� frequent unpredictable hypoglycaemia or� a marked dawn blood glucose increase.
NB. We find that in patients with unpredictable and wide swings in blood glucose levels during injection therapy, disablinghypoglycaemia may develop as injection therapy is intensified. These patients would then become candidates for CSII, asindicated above.
Prerequisites for insulin pump therapy
All patients should be:
� willing to undertake CSII� motivated� compliant in diabetes management� able to perform CSII procedures� able to perform frequent blood glucose self-monitoring� meet clinical indications for CSII� free of major psychological and psychiatric problems.
Patients who are likely to do poorly on pump therapy
� ‘brittle’ diabetic subjects characterised by recurrent DKA� patients who are poorly compliant on injection therapy� patients with significant psychological and psychiatric problems, or those who otherwise do not meet the prerequisites.
Source: Pickup and Keen, 2001.38
provided at a limited number of ‘area’ locations,this would disadvantage those living in rural areaswith poor public transport, particularly bearing inmind that they are more likely to be unable todrive than people without diabetes. One optionwould be for each area to start with one centre androll out the service according to demand.
Barriers to implementationUse of CSII in the UK appears to have beenlimited by two main factors. First, the cost haslimited use in those diabetes clinics whereconsultants believed in their value. Second, theprevailing clinical impression seems to have beenthat the risks might outweigh the benefits. Thatimpression probably dates from the days of theearly pumps, but has not been supersededperhaps because there has been little experiencewith newer and more reliable pumps. A circularscenario may have been operating: because thevalue of CSII was doubted by many diabetologists,there has been little pressure for more funds;because funds have been limited, there has beenlittle experience with newer pumps, and theconservative attitude has not been countered bynew experience.
Unpublished (as of early August 2003; Matsuoka K, Oxford: personal communication,2003) research from one English health authorityprovides very useful information on the barriers to
implementation if NICE approved the use ofinsulin pumps in selected individuals. Theproblems include:
� A lack of knowledge about CSII. � A perception amongst many in diabetes care
that the evidence base for CSII was weak; thisseemed to be based on knowledge of the earlytrials. The diabetologists with this belief did notthink that pumps had improved, or did notknow whether they had or not.
� Lack of resources, coupled with competingdemands, so that CSII would not be a toppriority even if funding became available.
� Concern about the training needs – it wasbelieved that heavy initial investment would berequired for education and backup in the firstfew weeks.
The three barriers reported most often were lackof a skills base (80%), insufficient staff for follow-up care (63%) and lack of funding. There wasparticular concern about a shortage of DSNs,which was not just about funding, but about theavailability of trained people.
This research project also noted that there was aninequitable distribution of pumps users acrosssocioeconomic areas, with very little use in themore deprived areas. This is only partly explainedby the fact that 50% of users fund their ownpumps.
Health Technology Assessment 2004; Vol. 8: No. 43
79
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
TABLE 42 Annual costs of Type 1 diabetes
All Type 1s (including children) (%) Number Annual cost (£)
1 3,300 3.5 million3 9,900 10.5 million5 16,500 17.5 million
This chapter reviews the key issues, and whereappropriate, the uncertainties involved in each.
EfficacyThe potential benefits of CSII are threefold:
1. Glycaemic control: the evidence suggests amodest improvement in control of high bloodglucose levels, as reflected in glycatedhaemoglobin. This may lead to reduced long-term complications, but evidence is lackingfrom the studies included in this review.
2. Frequency of hypoglycaemic episodes: althoughmost studies found no difference in thefrequency of hypoglycaemia between CSII andMDI, some found decreases with CSII and onestudy found an increase with CSII. There mayalso be benefits from reduced fear ofhypoglycaemia, but this has not beenquantified by the studies.
3. Well-being and QoL: QoL was assessed by justone eligible study, which reported nodifferences between CSII and MDI. Statementsfrom patients who are successful pump userssuggest an improvement in well-being andQoL, partly from greater flexibility of lifestyle.The effects have not been quantified in utilityterms.
There is good evidence that CSII is acceptable tousers (partly because those who do not like using itstop, and thereby incur no further added costs).The fact that many pump users are meeting thecosts themselves implies that they value it. Thereis also good evidence of a useful reduction inglycated haemoglobin levels and of a small dropin the frequency of serious hypoglycaemia.
There is strong anecdotal evidence from usersfrom INPUT and elsewhere that CSII improvesQoL, but the RCTs have neglected this and havefocused on measuring the more easily measurableoutcomes such as HbA1c. This is unfortunate: asone of our users commented, “there is more tocontrol of diabetes than HbA1”.
One problem with assessing the effect of CSII onglycated haemoglobin is that those who take part
in trials, or whose results are reported in someobservational studies, may be highly motivatedindividuals whose HbA1c is already better thanaverage, so that there is less scope forimprovement. Conversely, in some studies patientsmay have been poorly controlled with more togain, both in terms of glycated haemoglobin andreduction in hypoglycaemic events. For example,in the study by Boland and colleagues,98 the rateof severe hypoglycaemic episodes was reduced by50% (p = 0.01), and Bode and colleagues81 alsoreported much greater reductions inhypoglycaemia with CSII compared with MDI.Hence in routine care results may be better thanin the trials, because there is greater selection of patients with problems. Numbers of patientsgoing on to CSII may be small partly because the clinics that use CSII most are also vigorous in trying to achieve better control on MDI, so that only those who still have problems go on to CSII.
Two studies included in the meta-analysis byPickup and colleagues40 were excluded from the present meta-analysis. In the study by Marshall and colleagues,164 most patients (10 out of 12) used twice-daily injections while on the injection phase of the crossover study,and although the authors described the trial asbeing of “CSII versus optimised injectiontherapy”, it seemed more like optimisedconventional therapy rather than optimisedmodern MDI, and was therefore excluded.Because it was a very small study of only 10patients, inclusion would not have changed theresults of our meta-analysis.
The other study, by Helve and colleagues,165 wasagain of conventional therapy versus CSII.Although the authors say that conventionaltherapy could be optimised by changing thenumber and timing of daily injections, no data aregiven to show the extent to which this was done.They report that daily insulin dose did not changeduring the conventional phase (whereas itdropped during CSII), suggesting that in practicethere was little change in the intensity of theconventional treatment. Most patients were on twoinjections per day at the start, and a few were ononly one.
Health Technology Assessment 2004; Vol. 8: No. 43
81
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Chapter 7
Discussion: analysis of uncertainties
ComparatorsWe have assumed that patients progress fromconventional insulin treatment to MDI and then toCSII, and that the key comparator for CSII isoptimised MDI. However, given that there arevarying intensities of MDI, it is possible that interms of convenience to patients, changing to CSIImight be worthwhile for some patients before alloptions with MDI had been tried. That is, thatsome patients might be able to achieve goodcontrol, as reflected in HbA1c, on MDI but only ata level of inconvenience that made it worthwhileswitching to CSII before that level of control wasreached. This implies that patient conveniencebecomes an independent indicator for CSII. Theexpectation is that that would only apply topatients on MDI, but there could be some onconventional therapy (twice-daily mixtureinjections) who would request a change to CSII.
We do not have good data on the proportions ofpeople with Type 1 diabetes who achieve adequatecontrol at present. Data from young people withdiabetes in Scotland (SSGCYD) indicates that onlya minority succeed, but the proportion would beexpected to be higher in adults, especially theover-30s. Nevertheless, it may be worth bearing inmind that most people (perhaps 80%) with Type 1diabetes are not adequately controlled.166
There is a new, long-acting analogue insulin,glargine, on the market and another, detemir, isfollowing. Glargine has been the subject ofanother review.167 It may have advantages over thecurrent long-acting insulins,168 and might beregarded as a stronger competitor to CSII, byproviding a better (less hypoglycaemia) form ofMDI. However, CSII has the advantage of beingable to provide several different basal insulin ratesover 24 hours, whereas glargine provides only aconstant basal level. A recent abstract reportedthat CSII gave better glycaemic control thanglargine, but conclusions must await the full study,which was not an RCT.169
Cost-effectiveness and cost perQALYThere is general agreement that reducing HbA1c
improves long-term outcomes, and this has beenshown in the meta-analysis of smaller trials by Wangand colleagues18 (which pooled results from 681patients) and by the DCCT study19 (1441 patients).One problem in many of these studies is that theyreport changes in relative risks, rather than giving
absolute differences. However, they do provideevidence of delays in the onset of complicationssuch as renal disease, which should lead to life yearsgained in due course, although these may be20 years down the line and hence offset by the ruleson discounting. There should also be future savingsin healthcare, but these will be much more reducedby discounting at 6%. It may be worth noting thatthe main evidence to date from the DCCT was ofdelaying rather than preventing complications.
However, it has not proved possible to go fromreduction in HbA1c to a simple cost per QALY forseveral reasons. First, there is insufficient evidenceon the utility value of some of the benefits mostappreciated by patients, such as flexibility oflifestyle and freedom from the fear ofhypoglycaemic events.
Second, there is the issue of the improvement inHbA1c and the expected consequent reduction inthe risk of long-term complications. There areseveral problems with estimating the cost-effectiveness implications of these factors:
� Trials such as DCCT compare packages of care.The intensive arm of DCCT had CSII or MDI,more intensive self-monitoring of blood glucosewith targets to achieve, more education on dietand exercise, more visits to clinics and frequentadvice by telephone. It is unsafe to attribute allthe benefits seen in the intensive group to thereduction in HbA1c. For example, theircholesterol results were better; it may be thatsome benefits were due to aspects other thanthe insulin regimen. What other confoundingfactors are there in each trial? Did smokingrates or blood pressure levels change?
� There were clear differences in HbA1c levels.However, can we extrapolate from the benefitsof, a drop of, for example, 1.4% (from Wangand colleagues’ meta-analysis) to one of 0.6%(in the CSII analysis here)? If a drop of 1.4% inHbA1c gives an odds ratio of 0.32 for renaldisease progression, would the odds ratio for adrop of 0.65 be in proportion?
� There are differences in how these drops inHbA1c were obtained. Does a drop based on aswitch from conventional twice daily insulin toCSII have the same implication as one achievedby switching from MDI to CSII?
� There is also the issue of baseline HbA1c. Woulda drop from 10% to 9.4% carry the sameimplications as one from 8.0 to 7.4%? Even ifthe relative risk reduction was the same, thelower absolute risks would affect the cost-effectiveness equations.
Discussion: analysis of uncertainties
82
A third problem in producing a cost per QALY isthat we do not have good data on therepresentativeness of patients in the trials. Thosepatients who enter trials may do better whatevertreatment they use. If their results on MDI arebetter than average, there may be less scope forimprovement on CSII, which would reduce thepower of the studies to show benefits with CSII. Itmay be that there are other patients who wouldstruggle with an intensive MDI regimen but whowould cope with a pump. Pumps would be morecost-effective in the latter group because therewould be a greater potential gain in HbA1c.
A fourth problem might be that CSII is not just apump, but a package. However, since this alsoapplies to MDI, this should not invalidate themeta-analysis of CSII versus MDI.
CostsThe costs and uncertainties involved include thefollowing:
� The NHS costs of those with inadequate controlon MDI – will there be savings after transfer toCSII? Possibly, but probably small in the short-term. The cost to the NHS of hypoglycaemicevents is small in monetary terms.
� The cost of pumps. The number of pump usersin the UK is small at present. If a much greaternumber were to be used, could the NHS securesignificantly lower prices by bulk purchase?
� The cost of disposables such as infusion sets.Could these be reduced, perhaps by longer use?How much would that increase the risk ofinfections?
� The cost of insulin – good evidence of a modestbut useful reduction in the short term leadingto savings of around £90 per patient per annum
Duration of studiesMost of the present studies are fairly short term.We do not know what benefits will accrue after, forexample, 20 years of CSII.
Strengths and limitations ofreviewThe systematic review has the following strengths:
� The systematic review is independent of anyvested interests.
� The systematic review brings together theevidence for the effectiveness of insulin pumpsapplying consistent methods of criticalappraisal.
� The review was guided by the principles forundertaking systematic reviews. Beforeundertaking the review the methods were setout in research protocol (Appendix 1), whichwas commented on by an advisory group. Theprotocol defined the research question,inclusion criteria, quality criteria, dataextraction process and methods used toundertake the different stages of the review.
� An advisory group has informed the reviewfrom its initiation, through the development ofthe research protocol and completion of thereport.
In contrast, there were certain limitations placedupon the review:
� Owing to time constraints placed upon thereview, there was a lack of follow-up withauthors of studies to clarify methodologicaldetails and results from the primary studies.
� The systematic review was limited to parallelRCTs and randomised and non-randomisedcrossover studies. However, for areas where noeligible studies were identified, selectedobservational studies were discussed [see thesections ‘CSII in children’ (p. 37), ‘Overnight-only CSII in adults’ (p. 37), ‘Short-term CSII inpoorly controlled adults with Type 2 diabetes’(p. 37) and ‘Discontinuation rates’ (p. 42)].
� Abstracts and conference proceedings wereexcluded from the review as these usually fail toprovide adequate details of the methods of thestudy and their results, and inclusion waslimited to English language due to timeconstraints. A previous meta-analysis foundevidence of some publication bias for glycatedhaemoglobin.40
Other issues� Parallel RCTs and crossover studies have been
combined in meta-analyses in this systematicreview. In a recently published paper, Elbourneand colleagues suggest that combining crossoverand parallel trials in this way is conservative, asit ignores the within-patient correlation.170
� Due to the nature of the interventionscompared, blinding of patients and treatingclinicians in clinical trials is difficult, if notimpossible. Although a lack of blinding mayprovide the opportunity for observer bias, it wasfelt that as the primary outcome measureassessed was an objective measure, that of
Health Technology Assessment 2004; Vol. 8: No. 43
83
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
glycated haemoglobin, any potential bias wouldhave a limited effect.
� As part of a new development to help supportthe NICE appraisal process, the report includeda section examining the patients’ perspective onthe condition suffered and their reflections onthe different technologies assessed. Althoughconsidered useful, it should be recognised thatit is not methodologically rigorous and haslimitations that mean it is not generalisable.The information originates from members of aspecific user group, who are enthusiasts for thetechnology, may have paid for it themselves andhave had problems with previous forms ofmanaging the condition. Research is under wayto provide guidance on the most appropriatemethods for representing such patient views inhealth technology assessments.
Research needs1. We need much more evidence on the gains in
QoL from CSII, which was one of thehindrances in analysing cost-effectiveness. In particular, there is a need for trials
focused on patients with good control asrecorded by glycated haemoglobin, but withspecific problems such as recurrent severehypoglycaemia.
2. There is a need for good randomised trials inchildren, of different ages, and of differentpatterns of use, such as 24-hour use underparental supervision for the 0–4-years-olds,overnight only use in the 4–8-years age range and independent use in the over-8-years-olds.
3. Research is needed on the role of CSII inteenagers, who often have poor diabeticcontrol. Could pumps improve compliance withtreatment?
4. We need RCTs in pregnancy, startingpreconceptually, and compared with MDI.
5. There may be a place for short-term use ininsulin-resistant adults with Type 2 diabetes,but this is unproven. Trials should control forother interventions such as diet, and should bedone on outpatients.
6. There may be a need for further trials of CSIIversus MDI using longer acting analogues suchas glargine, but we need to see the full detailsof the current glargine trials first.
Discussion: analysis of uncertainties
84
Pump technology has advanced considerablysince the early days, and pumps are now
much more portable and reliable. In randomisedtrials, they show definite but modest benefits. In observational studies, greater benefits arereported, probably because the patients in those are selected because of particular problemsand have more scope to benefit. In routinepractice, patients who go on to pumps arecarefully selected and to a large degree self-selected.
However, pumps are used much less in the UKthan in other Western nations. This may be partlybecause of cost and partly because of early adverseexperiences. The lack of use may mean that theadverse experiences have in most centres not beenreplaced by positive ones, hence allowingpessimism to prevail.
It is unlikely that insulin pumps would be used bymore than a small percentage of people with Type1 diabetes, but the exact proportion is uncertain.
Health Technology Assessment 2004; Vol. 8: No. 43
85
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Chapter 8
Conclusions
We are grateful to the advisory panel whoprovided expert advice and comments on the
research protocol and/or a draft of this report:
Professor Stephanie Amiel (Professor of DiabeticMedicine), King’s College School of Medicine,London. Miss Ann Brooker (Paediatric DiabetesClinical Nurse Specialist), St Luke’s Hospital,Bradford. Mr John Davis, INPUT, Lymington,Hampshire. Sister Joan Everett (DiabetesSpecialist Nurse), Royal Bournemouth Hospital,Dorset. Dr Stephen Greene (Reader in Child andAdolescent Health ), Tayside Institute of ChildHealth, Ninewells Hospital and Medical School,Dundee. Professor Philip Home, Department ofDiabetes, The Medical School, Newcastle uponTyne. Dr Abigail King, Haverfordwest,Pembrokeshire. Dr John Pickup (Reader) Guy’s,King’s and St Thomas School of Medicine,London. Mrs Gill Salt (Paediatric Diabetes NurseSpecialist), New Cross Hospital, Wolverhampton.Dr Ken Tieszen, Department of Diabetes, HopeHospital, Salford.
We would like to thank the following people andorganisations for information or other assistance:Dr Pam Royle (Senior Researcher/InformationScientist), Wessex Institute of Health Research andDevelopment. Professor Harry Keen, Guy’sHospital, London. Karen Matsuoka, Oxford.Lorraine Rothery, Halifax. Professor WilliamTamborlane, School of Medicine, Yale University.Bournemouth Diabetes and Endocrine Centre.Diabetes Resource Centre, Harrogate DistrictHospital. INPUT, UK, and in particular all those
pump users who provided personal experiences.Royal College of Nursing Paediatric andAdolescent Diabetes group. Royal College ofPhysicians of London.
We are also grateful to Disetronic and Medtronicfor kindly supplying photographs of their insulinpumps.
This report was commissioned by the NHS R&DHTA Programme on behalf of NICE. The viewsexpressed in this report are those of the authorsand not necessarily those of the NHS R&D HTAProgramme. The report remains the responsibilityof the Southampton Health TechnologyAssessments Centre, Wessex Institute for HealthResearch and Development, University ofSouthampton. Any errors are the responsibility ofthe authors.
Contributions of authorsJL Colquitt (Senior Researcher) assessed studiesfor inclusion, extracted and synthesised clinicaleffectiveness data, and contributed to drafting thereport. C Green (Senior Research Fellow) wasresponsible for the cost-effectiveness section, andcontributed to drafting the report. MK Sidhu(Researcher) assisted with study selection, dataextraction and drafting the clinical effectivenesssection. D Hartwell (Researcher) assisted withstudy selection, data extraction and drafting thereport. N Waugh (Professor of Public Health)drafted the protocol, contributed to drafting thereport, and was responsible for writing thepatient’s perspective chapter.
Health Technology Assessment 2004; Vol. 8: No. 43
87
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Acknowledgements
1. Rangasami JJ, Greenwood DC, McSporran B,Smail PJ, Patterson CC, Waugh NR. Risingincidence of type 1 diabetes in Scottish children,1984–93. The Scottish Study Group for the Careof Young Diabetics. Arch Dis Child 1997;77:210–13.
2. Bingley PJ, Gale EA. Incidence of insulindependent diabetes in England: a study in theOxford region, 1985–6. BMJ 1989;298:558–60.
3. Zhao HX, Stenhouse E, Soper C, Hughes P,Sanderson E, Baumer JH, et al. Incidence ofchildhood-onset type 1 diabetes mellitus in Devonand Cornwall, England, 1975–1996. Diabet Med1999;16:1030–5.
4. Burden AC, Hearnshaw JR, Swift PG. Childhooddiabetes mellitus: an increasing incidence. DiabetMed 1989;6:334–6.
5. Patterson CC, Thorogood M, Smith PG, Heasman MA, Clarke JA, Mann JI. Epidemiologyof type 1 (insulin-dependent) diabetes in Scotland1968–1976: evidence of an increasing incidence.Diabetologia 1983;24:238–43.
6. Patterson CC, Smith PG, Webb J, Heasman MA,Mann JI. Geographical variation in the incidenceof diabetes mellitus in Scottish children during theperiod 1977–1983. Diabet Med 1988;5:160–5.
7. Staines A, Bodansky HJ, Lilley HE, Stephenson C,McNally RJ, Cartwright RA. The epidemiology ofdiabetes mellitus in the United Kingdom: theYorkshire Regional Childhood Diabetes Register.Diabetologia 1993;36:1282–7.
8. Green A, Gale EA, Patterson CC. Incidence ofchildhood-onset insulin-dependent diabetesmellitus: the EURODIAB ACE Study. Lancet1992;339:905–9.
9. UKPDS Study Group. Intensive blood glucosecontrol with sulphonylureas or insulin comparedwith conventional treatment and risk ofcomplications in patients with type 2 diabetes(UKPDS 33). Lancet 1998;352:837–53.
10. Neil HA, Gatling W, Mather HM. The OxfordCommunity Diabetes Study: evidence for anincrease in the prevalence of known diabetes inGreat Britain. Diabet Med 1987;4:539–43.
11. Gatling W, Budd S, Walters D, Mulle MA, Goddard JR, Hill RD. Evidence of an increasingprevalence of diagnosed diabetes in the Poole areafrom 1983 to 1996. Diabet Med 1998;15:1015–21.
12. Mather HM, Keen H. The Southall DiabetesSurvey: prevalence of known diabetes in Asiansand Europeans. BMJ 1985;291:1081–4.
13. Newnham A, Ryan R, Khunti K, Majeed A.Prevalence of diagnosed diabetes mellitus ingeneral practice in England and Wales, 1940 to1998. Health Stat Q 2002;14:5–13.
14. Turner R, Cull C, Holman R, for the UnitedKingdom Prospective Diabetes Study. UKPDS 17:A 9-year update of a randomised controlled trialof the effect of improved metabolic control oncomplications in non-insulin-dependant diabetesmellitus. Ann Intern Med 1996;124:136–45.
15. Waugh NR, Jung RT, Newton RW. The Dundeeprevalence study of insulin-treated diabetes;intervals between diagnosis and start of insulintherapy. Diabet Med 1989;6:346–50.
16. Humphrey ARG, McCarty DJ, Mackay IR, Rowley MJ, Dwyer T, Zimmet P. Autoantibodies toglutamic acid decarboxylase and phentotypicfeatures associated with early insulin treatment inindividuals with adult-onset diabetes. Diabet Med1998;15:113–19.
17. Office for National Statistics. Key health statisticsfrom general practice 1998. London: Office forNational Statistics; 2000.
18. Wang PH, Lau J, Chalmers TC. Meta-analysis ofeffects of intensive blood-glucose control on latecomplications of type I diabetes. Lancet1993;341:1306–9.
19. The Diabetes Control and Complications TrialResearch Group. The effect of intensive treatmentof diabetes on the development and progression oflong-term complications in insulin-dependentdiabetes mellitus. N Engl J Med 1993;329:977–86.
20. Reichard P, Nilsson B-Y, Rosenqvist U. The effectof long-term intensified insulin treatment on thedevelopment of microvascular complications ofdiabetes mellitus. N Engl J Med 1993;329:304–9.
21. The Diabetes Control and Complications Trial.The effect of intensive diabetes treatment on theprogression of diabetic retinopathy in insulin-dependent diabetes mellitus. Arch Ophthalmol1995;113:36–51.
22. Egger M, Davey-Smith G, Stettler C, Diem P. Risk of adverse effects of intensified treatment ininsulin-dependent diabetes mellitus: a meta-analysis. Diabet Med 1997;14:919–28.
Health Technology Assessment 2004; Vol. 8: No. 43
89
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
References
23. Gold AE, Frier BM. Hypoglycaemia: practical andclinical implications. In Kelnar CJH, editor.Childhood and adolescent diabetes, London: Chapmanand Hall; 1995. pp. 351–66.
24. The Diabetes Control and Complications TrialResearch Group. Hypoglycemia in the DiabetesControl and Complications Trial. Diabetes1997;46:271–86.
25. Muhlhauser I, Berger M, Sonnenberg G, Koch J,Jorgens V, Schernthaner G, et al. Incidence andmanagement of severe hypoglycemia in 434 adultswith insulin-dependent diabetes mellitus. DiabetesCare 1985;8:268–73.
26. Rewers A, Chase HP, Mackenzie T, Walravens P,Roback M, Rewers M, et al. Predictors of acutecomplications in children with type 1 diabetes.JAMA 2002;287:2511–18.
27. Emilien G, Maloteaux JM, Ponchon M.Pharmacological management of diabetes: recentprogress and future perspective in daily drugtreatment. Pharmacol Ther 1999;81:37–51.
28. Swift PGF. Insulin: types and regimens. Childhoodand adolescent diabetes. London: Chapman andHall; 1995. pp. 253–66.
29. Morris AD, Boyle DIR, McMahon AD, for theDARTS/MEMO Collaboration. Adherence toinsulin treatment, glycaemic control andketoacidosis in IDDM. Lancet 1997;350:1505–10.
30. Scottish Study Group for the Care of the YoungDiabetic. Factors influencing glycemic control inyoung people with type 1 diabetes in Scotland. Apopulation-based study (DIABAUD2). Diabetes Care2001;24:239–44.
31. Atkinson MA, Eisenbarth GS. Type 1 diabetes: new perspectives on disease pathogenesis andtreatment. Lancet 2001;358:221–9.
32. Pickup JC, Keen H, Parsons JA, Alberti KGMM.Continuous subcutaneous insulin infusion: anapproach to achieving normoglycaemia. BMJ1978;i:204–7.
33. Tamborlane WV, Sherwin RS, Genel M, Felig P.Reduction to normal of plasma glucose in juvenilediabetes by subcutaneous administration of insulinwith a portable infusion pump. N Engl J Med1979;300:573–8.
34. Thorp FK. Insulin pump therapy reconsidered.JAMA 1986;255:645–6.
35. Davies AG, Baum JD. A decade of insulin infusionpumps. Arch Dis Child 1988;63:329–32.
36. Lenhard MJ, Reeves GD. Continuoussubcutaneous insulin infusion: a comprehensivereview of insulin pump therapy. Arch Intern Med2001;161:2293–300.
37. American Diabetes Association. Continuoussubcutaneous insulin infusion. Diabetes Care2002;25:S116.
38. Pickup J, Keen H. Continuous subcutaneousinsulin infusion in type 1 diabetes is beneficial inselected patients and should be more widelyavailable. BMJ 2001;322:1263–4.
39. Pickup J, Keen H. Continuous subcutaneousinsulin infusion at 25 years: Evidence base for theexpanding use of insulin pump therapy in type 1diabetes. Diabetes Care 2002;25:593–8.
40. Pickup J, Mattock M, Kerry S. Glycaemic controlwith continuous subcutaneous insulin infusioncompared with intensive insulin injections inpatients with type 1 diabetes: meta-analysis ofrandomised controlled trials. BMJ 2002;324:705–8.
41. Pons JMV. Continuous subcutaneous infusion of insulinwith portable pump in diabetes type 1 patients.Barcelona: CAHTA; 2000. No. IN01/2000.
42. Munoz A, Saz Z, Amate JM. Efficacy of insulininfusion pumps. Impact on the quality of live of certainpatients. Spain: AETS; 2000.
43. Hotta SS, Adams D. Reassessment of externalinsulin infusion pumps. Health Technol AssessReports 1990;(9):1–9.
44. Graff MR, Rubin RR, Walker EA. How diabetesspecialists treat their own diabetes: findings from astudy of the AADE and ADA membership. DiabetesEduc 2000;26:460–7.
45. Wredling R, Lin PE, Adamson U. Pump ‘run-away’causing severe hypoglycaemia. Lancet 1989;ii:273.
46. Tritos NA, Casper K, King G. Museum visitleading to insulin pump malfunction. Ann InternMed 1997;126:746.
47. Darmon P, Guillaume V, Wiart J, Dutour A, Oliver C. Do mobile cellular phones interfere withportable insulin pumps? Diabetes Care 1998;21:1775.
48. Prendergast JJ, Dorsey C, Elsea V. Over-delivery ofinsulin by insulin pumps. Diabetes Care 1995;18:1201–2.
49. Mecklenburg RS, Benson EA, Benson JW,Fredlund PN, Guinn T, Metz RJ, et al. Acutecomplications associated with insulin infusionpump therapy. Report of experience with 161patients. JAMA 1984;252:3265–9.
50. Teutsch SM, Herman WH, Dwyer DM, Lane JM.Mortality among diabetic patients usingcontinuous subcutaneous insulin-infusion pumps.N Engl J Med 1984;310:361–8.
51. Knight G, Jennings AM, Boulton AJ, Tomlinson S,Ward JD. Severe hyperkalaemia and ketoacidosisduring routine treatment with an insulin pump.BMJ 1985;291:371–2.
52. The Kroc Collaborative Study Group. Bloodglucose control and the evolution of diabeticretinopathy and albuminuria. A preliminarymulticenter trial. N Engl J Med 1984;311:365–72.
References
90
53. Steel JM, West CP. Intrauterine death duringcontinuous subcutaneous infusion of insulin. BMJ1985;290:1787.
54. Bradley C, Gamsu DS, Knight G, Boulton AJ,Ward JD. Predicting risk of diabetic ketoacidosis inpatients using continuous subcutaneous insulininfusion. BMJ 1986;293:242–3.
55. Bending JJ, Pickup JC, Keen H. Frequency ofdiabetic ketoacidosis and hypoglycemic comaduring treatment with continuous subcutaneousinsulin infusion. Audit of medical care. Am J Med1985;79:685–91.
56. Chantelau E, Spraul N, Mhuhlhauser I, Gause R,Beger M. Long-term safety, efficacy and side-effects of continuous subcutaneous insulin infusiontreatment for type 1 (insulin-dependent) diabetesmellitus: a one centre experience. Diabetologia1989;32:421–6.
57. Raskin P. Treatment of type 1 diabetes withportable insulin pumps. Diabetes Care 1982;5(Suppl 1):48–52.
58. Toth EL, Boychuk LR, Kirkland PA. Recurrentinfection of continuous subcutaneous insulininfusion sites with Mycobacterium fortuitum. DiabetesCare 1995;18:1284–5.
59. Pagnoux C, Nassif X, Boitard C, Timsit J.Infection of continuous subcutaneous insulininfusion site with Mycobacterium peregrinum. DiabetesCare 1998;21:191–2.
60. New insulin pump forum launched in UK. Pract Diab Int 2000;17:278–9.
61. Brink SJ, Stewart C. Insulin pump treatment ininsulin-dependent diabetes mellitus. Children,adolescents, and young adults. JAMA 1986;255:617–21.
62. Levy-Marchal C, Czernichow P. Feasibility ofcontinuous subcutaneous insulin infusion in youngdiabetic patients. Diabetes Metab 1988;14:108–13.
63. Knight G, Boulton AJ, Ward JD. Experience ofcontinuous subcutaneous insulin infusion in theoutpatient management of diabetic teenagers.Diabet Med 1986;3:82–4.
64. Bougneres PF, Landier F, Lemmel C, Mensire A,Chaussain JL. Insulin pump therapy in youngchildren with type 1 diabetes. J Pediatr 1984;105:212–7.
65. De Beaufort CE, Bruining GJ. Continuoussubcutaneous insulin infusion in children. Diabet Med 1987;4:103–8.
66. Kaufman FR, Halvorson M, Kim C,Pitukcheewanont P. Use of insulin pump therapyat nighttime only for children 7–10 years of agewith type 1 diabetes. Diabetes Care 2000;23:579–82.
67. Schiffrin A. Nighttime continuous subcutaneousinsulin infusion revisited: a strategy for improvinginsulin delivery. Diabetes Care 2000;23:571–3.
68. Olsson P, Arnqvist H, Von Schenck H, Ottosson AM. Overnight metabolic control withbedtime injection of intermediate-acting insulin orcontinuous subcutaneous insulin fusion. DiabetesCare 1987;10:702–6.
69. Kanc K, Janssen MM, Keulen ET, Jacobs MA,Popp-Snijders C, Snoek FJ, et al. Substitution ofnight-time continuous subcutaneous insulininfusion therapy for bedtime NPH insulin in amultiple injection regimen improvescounterregulatory hormonal responses andwarning symptoms of hypoglycaemia in IDDM.Diabetologia 1998;41:322–9.
70. Gabbe SG. New concepts and applications in theuse of the insulin pump during pregnancy. J Matern Fetal Med 2000;9:42–5.
71. Aucott SW, Williams TG, Hertz RH, Kalhan SC.Rigorous management of insulin-dependentdiabetes mellitus during pregnancy. Acta Diabetol1994;31:126–9.
72. Caruso A, Lanzone A, Bianchi V, Massidda M,Castelli MP, Fulghesu AM, et al. Continuoussubcutaneous insulin infusion (CSII) in pregnantdiabetic patients. Prenat Diagn 1987;7:41–50.
73. Jensen BM, Kuhl C, Molsted-Pedersen L, Saurbrey N, Fog-Pedersen J. Preconceptionaltreatment with insulin infusion pumps in insulin-dependent diabetic women with particularreference to prevention of congenitalmalformations. Acta Endocrinol Suppl (Copenh)1986;277:81–5.
74. Rudolf MC, Coustan DR, Sherwin RS, Bates SE,Felig P, Genel M, et al. Efficacy of the insulin pumpin the home treatment of pregnant diabetics.Diabetes 1981;30:891–5.
75. Simmons D, Thompson CF, Conroy C, Scott DJ.Use of insulin pumps in pregnancies complicatedby Type 2 diabetes and gestational diabetes in amultiethnic community. Diabetes Care2001;24:2078–82.
76. Brinchmann-Hansen O, Dahl-Jorgensen K,Hanssen KF, Sandvik L. The response of diabeticretinopathy to 41 months of multiple insulininjections, insulin pumps, and conventional insulintherapy. Arch Ophthalmol 1988;106:1242–6.
77. Dahl-Jorgensen K, Hanssen KF, Kierulf P, Bjoro T,Sandvik L, Aagenaes O. Reduction of urinaryalbumin excretion after 4 years of continuoussubcutaneous insulin infusion in insulin-dependent diabetes mellitus. The Oslo Study. ActaEndocrinologica 1988;117:19–25.
78. Dahl-Jorgensen K, Brinchmann-Hansen O,Hanssen KF, Ganes T, Kierulf P, Smeland E, et al.Effect of near normoglycaemia for two years onprogression of early diabetic retinopathy,nephropathy, and neuropathy: the Oslo study. BMJ1986;293:1195–9.
Health Technology Assessment 2004; Vol. 8: No. 43
91
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
79. Tsui E, Barnie A, Ross S, Parkes R, Zinman B.Intensive insulin therapy with insulin lispro: arandomized trial of continuous subcutaneousinsulin infusion versus multiple daily insulininjection. Diabetes Care 2001;24:1722–7.
80. Hanaire-Broutin H, Melki V, Bessieres-Lacombe S,Tauber JP. Comparison of continuoussubcutaneous insulin infusion and multiple dailyinjection regimens using insulin lispro in type 1diabetic patients on intensified treatment: arandomized study. The Study Group for theDevelopment of Pump Therapy in Diabetes.Diabetes Care 2000;23:1232–5.
81. Bode BW, Steed RD, Davidson PC. Reduction insevere hypoglycemia with long-term continuoussubcutaneous insulin infusion in type I diabetes.Diabetes Care 1996;19:324–7.
82. Chiasson JL, Ducros F, Poliquin HM, Lopez D,Lecavalier L, Hamet P. Continuous subcutaneousinsulin infusion (Mill-Hill Infuser) versus multipleinjections (Medi-Jector) in the treatment ofinsulin-dependent diabetes mellitus and the effectof metabolic control on microangiopathy. DiabetesCare 1984;7:331–7.
83. Haakens K, Hanssen KF, Dahl-Jorgensen K, Vaaler S, Aagenaes O, Mosand R. Continuoussubcutaneous insulin infusion (CSII), multipleinjections (MI) and conventional insulin therapy(CT) in self-selecting insulin-dependent diabeticpatients. A comparison of metabolic control, acutecomplications and patient preferences. J InternMed 1990;228:457–64.
84. Ziegler D, Dannehl K, Koschinsky T, Toeller M,Gries FA. Comparison of continuous subcutaneousinsulin infusion and intensified conventionaltherapy in the treatment of type I diabetes: a two-year randomized study. Diabetes Nutr Metab ClinExp 1990;3:203–13.
85. Nosadini R, Velussi M, Fioretto P, Doria A,Avogaro A, Trevisan R, et al. Frequency ofhypoglycaemic and hyperglycaemic–ketoticepisodes during conventional and subcutaneouscontinuous insulin infusion therapy in IDDM.Diabetes Nutr Metab 1988;1:289–96.
86. Schmitz A, Christiansen JS, Christensen CK,Hermansen K, Mogensen CE. Effect of pumpversus pen treatment on glycaemic control andkidney function in long-term uncomplicatedinsulin-dependent diabetes mellitus (IDDM). DanMed Bull 1989;36:176–8.
87. Bak JF, Nielsen OH, Pedersen O, Beck-Nielsen H.Multiple insulin injections using a pen injectorversus insulin pump treatment in young diabeticpatients. Diabetes Res 1987;6:155–8.
88. Home PD, Capaldo B, Burrin JM, Worth R,Alberti KG. A crossover comparison of continuoussubcutaneous insulin infusion (CSII) against
multiple insulin injections in insulin-dependentdiabetic subjects: improved control with CSII.Diabetes Care 1982;5:466–71.
89. Nathan DM, Lou PAJ. Intensive conventional andinsulin pump therapy in adult type 1 diabetes. A crossover study. Ann Intern Med 1982;97:31–6.
90. Saurbrey N, Arnold-Larsen S, Moller-Jensen B,Kuhl C. Comparison of continuous subcutaneousinsulin infusion with multiple insulin injectionsusing the NovoPen. Diabet Med 1988;5:150–3.
91. Schiffrin A, Belmonte MM. Comparison betweencontinuous subcutaneous insulin infusion andmultiple injection of insulin. A one-yearprospective study. Diabetes 1982;31:255–64.
92. Nosari I, Maglio ML, Lepore G. Is continuoussubcutaneous insulin infusion more effective thanintensive conventional insulin therapy in thetreatment of pregnant diabetic women? DiabetesNutr Metab 1993;6:33–7.
93. Coustan DR, Reece EA, Sherwin RS, Rudolf MC,Bates SE, Sockin SM, et al. A randomized clinicaltrial of the insulin pump vs intensive conventionaltherapy in diabetic pregnancies. JAMA1986;255:631–6.
94. Carta Q, Meriggi E, Trossarelli GF, Catella G, Dal Molin V, Menato G, et al. Continuoussubcutaneous insulin infusion versus intensiveconventional insulin therapy in type I and type IIdiabetic pregnancy. Diabetes Metab 1986;12:121–9.
95. Burkart W, Hanker JP, Schneider HP.Complications and fetal outcome in diabeticpregnancy. Intensified conventional versus insulinpump therapy. Gynecol Obstet Invest 1988;26:104–12.
96. Tamborlane WV, Batas SE, Rudolf MC.Comparison of continuous subcutaneous insulininfusion versus multiple daily injections inadolescents with insulin-dependent diabetes. Adv Diabetol 1989;2 (Suppl 1):24–7.
97. Schiffrin AD, Desrosiers M, Aleyassine H,Belmonte MM. Intensified insulin therapy in thetype 1 diabetic adolescent: a controlled trial.Diabetes Care 1984;7:107–13.
98. Boland EA, Grey M, Oesterle A, Fredrickson L,Tamborlane WV. Continuous subcutaneous insulininfusion. A new way to lower risk of severehypoglycemia, improve metabolic control, andenhance coping in adolescents with type 1diabetes. Diabetes Care 1999;22:1779–84.
99. Mack-Fogg J, Bates S, Faro B, Sciera M, Ippolito K, Orlowski CC, et al. Safe and effectiveuse of continuous subcutaneous insulin infusion inyoung children with type 1 diabetes mellitus.Diabetes 2002;51 (Suppl 1, poster 712):122A.
100. Rudolf JW, Hirsch IB. An assessment ofcontinuous subcutaneous insulin infusion therapy
References
92
in an academic diabetes clinic. Diabetes 2000;49 (Suppl 1, poster 501-P):A124.
101. Haardt MJ, Berne C, Dorange C, Slama G, Selam JL. Efficacy and indications of CSII revisited:the Hotel-Dieu cohort. Diabet Med 1997;14:407–8.
102. Tamborlane WV, Bonfig W, Boland E. Recentadvances in treatment of youth with type 1diabetes: better care through technology. DiabetMed 2001;18:864–70.
103. Kaufman FR, Halvorson M, Miller D, Mackenzie M, Fisher LK, Pitukcheewanont P.Insulin pump therapy in type 1 pediatric patients:now and into the year 2000. Diabetes Metab Res Rev1999;15:338–52.
104. Plotnick L, Clark L. Insulin pumps in children andadolescents. Endocrinologist 2001;11:112–17.
105. Steindel BS, Roe TR, Costin G, Carlson M,Kaufman FR. Continuous subcutaneous insulininfusion (CSII) in children and adolescents withchronic poorly controlled type 1 diabetes mellitus.Diabetes Res Clin Pract 1995;27:199–204.
106. Maniatis AK, Klingensmith GJ, Slover RH, Mowry CJ, Chase HP. Continuous subcutaneousinsulin infusion therapy for children andadolescents: an option for routine diabetes care.Pediatrics 2001;107:351–6.
107. Conrad SC, McGrath MT, Gitelman SE. Transition from multiple daily injections tocontinuous subcutaneous insulin infusion in type 1diabetes mellitus. J Pediatr 2002;140:235–40.
108. Ahern JA, Boland E, Doane R, Ahern JJ, Rose P,Vincent M, et al. Insulin pump therapy inpediatrics: a therapeutic alternative to safely lowerHbA1c levels across all age groups. Pediatr Diabetes2002;3:10–15.
109. Fox LA, Wilkinson K, Buckloh L, Wysocki T,Mauras N. A randomized trial of insulin pumptherapy in preschool age children with type 1diabetes mellitus: preliminary results. Diabetes2002;51 (Suppl 2, poster 1751):426A.
110. Weintrob N, Benzaquen H, Galatzer A, Shalitin S,Lazar L, Fayman G, et al. Comparison ofcontinuous subcutaneous insulin infusion andmultiple daily injection regimens in children withtype 1 diabetes: a randomised controlled trial.Diabetes 2002;51 (Suppl 2, poster 1969).
111. Valensi P, Moura I, Le Magoarou M, Paries J,Perret G, Attali JR. Short-term effects ofcontinuous subcutaneous insulin infusiontreatment on insulin secretion in non-insulin-dependent overweight patients with poorglycaemic control despite maximal oral anti-diabetic treatment. Diabetes Metab 1997;23:51–7.
112. Dupuy O, Mayaudon H, Palou M, Sarret D,Bordier L, Bauduceau B. Optimized transientinsulin infusion in uncontrolled type 2 diabetes:
evaluation of a pragmatic attitude. Diabetes Metab2000;26:371–5.
113. Bode B, Weinstein R, Bell D, McGill J, Nadeau D,Raskin P, et al. Comparison of insulin aspart withbuffered regular insulin and insulin lispro incontinuous subcutaneous insulin infusion: arandomized study in type 1 diabetes. Diabetes Care2002;25:439–44.
114. Renner R, Pfutzner A, Trautmann M, Harzer O,Sauter K, Landgraf R. Use of insulin lispro incontinuous subcutaneous insulin infusiontreatment. Results of a multicenter trial. GermanHumalog-CSII Study Group. Diabetes Care 1999;22:784–8.
115. Schmauss S, Konig A, Landgraf R. Human insulinanalogue [LYS(B28), PRO(B29)]: the ideal pumpinsulin? Diabet Med 1998;15:247–9.
116. Melki V, Renard E, Lassmann-Vague V, Boivin S,Guerci B, Hanaire-Broutin H, et al. Improvementof HbA1c and blood glucose stability in IDDMpatients treated with lispro insulin analog inexternal pumps. Diabetes Care 1998;21:977–82.
117. Raskin P, Holcombe JH, Tamborlane WV, MaloneJI, Strowig S, Ahern JA, et al. A comparison ofinsulin lispro and buffered regular human insulinadministered via continuous subcutaneous insulininfusion pump. J Diabetes Complications 2001;15:295–300.
118. Zinman B, Tildesley H, Chiasson JL, Tsui E,Strack T. Insulin lispro in CSII: results of a double-blind crossover study. Diabetes 1997;46:440–3.
119. Green S, Deeks J, Savio F. The ends and themeans: continuous data in Cochrane reviews.Poster presentation at Cochrane Colloquium,2001.
120. Schifferdecker E, Schmidt K, Boehm BO, Schatz H.Long-term compliance of intensified insulintherapy. Diabetes Res Clin Pract 1994;23:17–23.
121. Wredling R, Lins PE, Adamson U. Factorsinfluencing the clinical outcome of continuoussubcutaneous insulin infusion in routine practice.Diabetes Res Clin Pract 1993;19:59–67.
122. Knight G, Boulton AJ, Drury J, Gamsu DS, Moses JL, Bradley C, et al. A feasibility study ofthe use of continuous subcutaneous insulininfusion in a diabetic clinic: patients’ choice oftreatment. Diabet Med 1984;1:267–72.
123. Floyd JC, Cornell RG, Jacober SJ, Griffith LE,Funnell MM, Wolf LL, et al. A prospective studyidentifying risk factors for discontinuance ofinsulin pump therapy. Diabetes Care 1993;16:1470–8.
124. Bell DS, Ackerson C, Cutter G, Clements RS.Factors associated with discontinuation ofcontinuous subcutaneous insulin infusion. Am JMed Sci 1988;295:23–8.
Health Technology Assessment 2004; Vol. 8: No. 43
93
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
125. Bell DSH, Cutter G, Clements J. The feasibility oflong-term treatment of diabetes with continuoussubcutaneous insulin infusion. Diabetes Nutr MetabClin Exp 1993;6:57–60.
126. British National Formulary. London: BritishNational Formulary; 2002.
127. Herman WH, Eastman RC. The effects oftreatment on the direct costs of diabetes. DiabetesCare 1998;21:C19–24.
128. American Diabetes Association. Economicconsequences of diabetes mellitus in the U.S. in1997. Diabetes Care 1998;21:296–309.
129. Lifetime benefits and costs of intensive therapy aspracticed in the diabetes control andcomplications trial. The Diabetes Control andComplications Trial Research Group. JAMA1996;276:1409–15.
130. Gray A, Raikou M, McGuire A, Fenn P, Stevens R,Cull C, et al. Cost effectiveness of an intensiveblood glucose control policy in patients with type 2diabetes: economic analysis alongside randomisedcontrolled trial (UKPDS 41). United KingdomProspective Diabetes Study Group. BMJ2000;320:1373–8.
131. Stratton IM, Adler AI, Neil HA, Matthews DR,Manley SE, Cull CA, et al. Association of glycaemiawith macrovascular and microvascularcomplications of type 2 diabetes (UKPDS 35):prospective observational study. BMJ2000;321:405–12.
132. Klein R, Klein BE, Moss SE, Cruickshanks KJ. The Wisconsin Epidemiologic Study of DiabeticRetinopathy: XVII. The 14-year incidence andprogression of diabetic retinopathy and associatedrisk factors in type 1 diabetes. Ophthalmology1998;105:1801–15.
133. Effect of intensive blood-glucose control withmetformin on complications in overweightpatients with type 2 diabetes (UKPDS 34). UKProspective Diabetes Study (UKPDS) Group.Lancet 1998;352:854–65.
134. Clark P, Gray A, Adler A, Stevens R, Raikou M,Cull C, et al. Cost-effective analysis of intensiveblood glucose control with metformin inoverweight patients with type 2 diabetes (UKPDS51). Diabetologia 2001;44:298–304.
135. Palmer AJ, Weiss C, Sendi PP, Neeser K, Brandt A,Singh G, et al. The cost-effectiveness of differentmanagement strategies for type I diabetes: a Swissperspective. Diabetologia 2000;43:13–26.
136. INPUT. INsulin PUmp Therapy. 2001.
137. Directory of diabetes care. Cambridge: CMA MedicalData; 2001.
138. Netten A, Rees T, Harrison G. Unit costs of healthand social care, 2001. Canterbury: Personal Social
Services Research Unit, University of Kent atCanterbury; 2002.
139. NHS. NHS reference costs. London: Department ofHealth; 2000.
140. Nordfeldt S, Jonsson D. Short-term effects ofsevere hypoglycaemia in children and adolescentswith type 1 diabetes. A cost-of-illness study. ActaPaediatr 2001;90:137–42.
141. Leese GP, Wang J, Broomhall J, Kelly P, Marsden A,Morrison W, et al. Incidence of severehypoglycaemia requiring emergency treatment intype 1 and type 2 diabetes: a population-basedstudy of health service resource use. Diabet Med2002;19 (Suppl 2):12–3.
142. Rubin RR, Peyrot M. Quality of life and diabetes.Diabetes Metab Res Rev 1999;15:205–18.
143. The Diabetes Control and Complications TrialResearch Group. The relationship of glycemicexposure (HbA1c) to the risk of development andprogression of retinopathy in the diabetes controland complications trial. Diabetes 1995;44:968–83.
144. Klein R, Klein BE, Moss SE, Davis MD, DeMets DL.Glycosylated hemoglobin predicts the incidenceand progression of diabetic retinopathy. JAMA1988;260:2864–71.
145. Klein R, Klein BE. Relation of glycemic control todiabetic complications and health outcomes.Diabetes Care 1998;21 (Suppl 3):C39–43.
146. Epidemiology of severe hypoglycemia in thediabetes control and complications trial. TheDCCT Research Group. Am J Med 1991;90:450–9.
147. Microvascular and acute complications in IDDMpatients: the EURODIAB IDDM ComplicationsStudy. Diabetologia 1994;37:278–85.
148. Influence of intensive diabetes treatment onquality-of-life outcomes in the diabetes control andcomplications trial. Diabetes Care 1996;19:195–203.
149. Ferguson SC, Strachan MW, Janes JM, Frier BM.Severe hypoglycaemia in patients with type 1diabetes and impaired awareness ofhypoglycaemia: a comparative study of insulinlispro and regular human insulin. Diabetes MetabRes Rev 2001;17:285–91.
150. Marrero DG, Guare JC, Vandagriff JL, Fineberg NS.Fear of hypoglycemia in the parents of childrenand adolescents with diabetes: maladaptive orhealthy response? Diabetes Educ 1997;23:281–6.
151. Cox DJ, Irvine A, Gonder-Frederick L, NowacekG, Butterfield J. Fear of hypoglycaemia:quantification, validation and utilisation. DiabetesCare 1987;10:6717–621.
152. Clarke WL, Cox DJ, Gonder-Frederick LA, Julian D, Schlundt D, Polonsky W. Reducedawareness of hypoglycemia in adults with IDDM. Aprospective study of hypoglycemic frequency and
References
94
associated symptoms. Diabetes Care 1995;18:517–22.
153. Gold AE, MacLeod KM, Frier BM. Frequency ofsevere hypoglycaemia in patients with type 1diabetes with impaired awareness ofhypoglycaemia. Diabetes Care 1994;17:697–703.
154. Chantelau E, Schiffers T, Schutze J, Hansen B.Effect of patient-selected intensive insulin therapyon quality of life. Patient Educ Couns1997;30:167–73.
155. Lewis KS, Bradley C, Knight G, Boulton AJM,Ward JD. A measure of treatment satisfactiondesigned specifically for people with insulin-dependent diabetes. Diabet Med 1988;5:235–42.
156. Johansson UB, Adamson UC, Lins PE, Wredling RA. Improved blood glucose variability,HbA1c insuman Infusat and less insulinrequirement in IDDM patients using insulin lisproin CSII. The Swedish Multicenter Lispro InsulinStudy. Diabetes Metab 2000;26:192–6.
157. Wu S, Sainfort F, Tomar R, Tollios JL, Fryback DG.Development and application of a model toestimate the impact of type 1 diabetes on health-related quality of life. Diabetes Care 1998;21:725–31.
158. Brown MM, Brown GC, Sharma S, Shah G. Utilityvalues and diabetic retinopathy. Am J Ophthalmol1999;128:324–30.
159. Kiberd BA, Jindal KK. Screening to prevent renalfailure in insulin dependent diabetic patients: aneconomic evaluation. BMJ 1995;311:1595–9.
160. Tomar R, Lee S, Wu S, Klein R, Moss SE, Fryback DG, et al. Disease progression and cost ofinsulin dependent diabetes mellitus: developmentand application of simulation model. J Soc HealthSyst 1998;5:24–37.
161. Dolan P. Modelling valuations for EuroQol healthstates. Med Care 1997;35:1095–108.
162. Gerich JE. Selection of patients for intensiveinsulin therapy. Arch Intern Med 1985;145:1383–4.
163. Hanas R. Selection for and initiation of continuoussubcutaneous insulin infusion. Proceedings from aworkshop. Horm Res 2002;57 (Suppl 1):101–4.
164. Marshall SM, Home PD, Taylor R, Alberti KG.Continuous subcutaneous insulin infusion versusinjection therapy: a randomised cross-over trialunder usual diabetic clinic conditions. Diabet Med1987;4:521–5.
165. Helve E, Koivisto VA, Lehtonen A, Pelkonen R,Huttunen JK, Nikkila EA. A crossover comparisonof continuous insulin infusion and conventionalinjection treatment of type I diabetes. Acta MedScand 1987;221:385–93.
166. Scottish Study Group for the Care of the YoungDiabetic. Factors influencing glycemic control inyoung people with type 1 diabetes in Scotland.Diabetes Care 2001;24:239–44.
167. Warren E, Weatherley-Jones E, Chilcott J, Beverley C. Systematic review and economicevaluation of a long-acting insulin analogue,insulin glargine. Health Technol Assess 2004;8(in press).
168. Schober E, Schoenle E, van Dyk J, Wernike-Panten K. Comparative trial betweeninsulin glargine and NPH insulin in children andadolescents with type 1 diabetes. Diabetes Care2001;24:2005–6.
169. Armstrong DU, King AB. Basal insulin: continuousglucose monitoring reveals less overnighthypoglycaemia with continuous subcutaneousinsulin infusion than with glargine. Diabetes2002;51 (Suppl 2, poster 373):92A.
170. Elbourne DR, Altman DG, Higgins JPT, Curtin F,Worthington HV, Vail A. Meta-analyses involvingcross-over trials: methodological issues. Int JEpidemiol 2002;31:140–9.
171. Chilcott J, Wight J, Lloyd Jones M, Tappenden P.The clinical effectiveness and cost effectiveness ofpioglitazone in Type 2 diabetes mellitus: a rapidand systematic review. Health Technol Assess2001;5(19).
Health Technology Assessment 2004; Vol. 8: No. 43
95
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Full title of research questionResearch aim: to assess the clinical effectivenessand cost-effectiveness of continuous subcutaneousinsulin infusion (CSII) using insulin pumps in thetreatment of patients with insulin-treated diabetes,whether Type 1 or Type 2, in whom diabetescontrol is unsatisfactory, and to consider whichgroups of patients could benefit most.
Clarification of research questionand scopeConventional insulin treatment in people withType 1 diabetes involves two injections per day,usually giving a mixture of short- and longeracting insulins [this will be referred to henceforthas conventional therapy (CT)]. CT providessatisfactory control as assessed by glycosylatedhaemoglobin (HbA1c) in only a minority ofpatients. Furthermore, some of those who doachieve an acceptable average blood glucose levelonly do so by having a mixture of high and lowglucose levels, and may have lives made difficultby hypoglycaemic attacks. The HbA1c level cantherefore be used only as a guide and needs to beused in conjunction with information onfluctuations in blood glucose, such as number ofhypoglycaemic episodes, and results of blood tests,in order to establish how the HbA1c level has beenachieved.
After CT, the current next best regimen is ‘basalplus bolus’, now often referred to as ‘meal-timeplus basal’, wherein patients receive one or twoinjections of long-acting insulins to provide thelow level of insulin needed throughout the dayand night, with additional injections of short-acting insulins to cover the steep rises in bloodglucose that would otherwise occur after meals.This is sometimes referred to as MDI, for multipledaily injections, and can be regarded as a form ofintensified treatment. However, there can bedifferent degrees of intensity. Our initial searchesshow variations in the meanings of conventionaland intensive. The term ‘optimised MDI’ has beenused to describe the situation where control withMDI is thought to be as good as can be achieved.This may involve four or more injections each day,
usually accompanied by frequent self-testing ofblood sugar levels using reagent strips and meters.
CSII is usually used to provide intensive treatmentin a different way to MDI, with a needle under theskin all day long, but with different amounts ofinsulin being given at different times – a slowbasal infusion all day long, but with the infusionrate being boosted to cover meals. It is thereforeanother way of providing basal plus bolus, but hasadvantages in that both basal and bolus dosagecan be adjusted more easily and that because only short-acting insulin is used, there is lesschance of hypoglycaemic episodes fromunpredictable absorption from injections of long-acting insulins. There can also be different basalrates at different times of day.
The main clinical question is the extent to whichCSII using insulin pumps provides any clinicaladvantages over management of Type 1 diabeteswith optimised MDI. The benefits could be bettercontrol of blood glucose as reflected in HbA1, or asimilar level of control but with other advantagessuch as fewer problems with hypoglycaemia, orgreater flexibility of lifestyle and hence betterQoL. If MDI is particularly intensive (e.g.5–7 injections a day), then CSII may in effect beless intensive.
Multiple injections can be given by syringe ormore often by insulin pen injectors or acombination thereof.
It is expected that CSII will be part of a packageof care delivered by specialist clinics.
The cost-effectiveness question is whether thebenefits are commensurate with the marginal cost.
Subsidiary questions to be addressed include:
� Do rapid-acting analogue insulins haveadvantages over older short-acting (soluble)insulins?
� Should CSII be used for women with Type 1diabetes preconceptually and during pregnancy?
Where possible, cost issues will be considered forthese questions.
Health Technology Assessment 2004; Vol. 8: No. 43
97
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Appendix 1
Rapid review methods from the research protocol
Studies giving data on discontinuation rates will beused as a guide to patient acceptability and otherdata on patient experiences and perspective willbe obtained from users and past users.
Studies on safety with modern pumps will besought. With the older pumps in the past, therewere problems with device failure leading to DKA.
Two other possible questions have been identifiedduring initial literature searches, but where littleresearch seems to have been done. These will onlybe addressed if time, resources and evidencepermit. Our initial impression is that the presentevidence will be insufficient as a basis for guidanceand that these may be topics to be identified onlyas areas for future research. They are:
� What is the value of short periods of CSII inpatients with Type 2 diabetes who are veryresistant to oral drugs?
� What is the role of overnight-only CSII (inpatients who would take injections as usual byday)? This method might be particularlyrelevant to children, but could apply to someadults as well.
These questions will not be subjected to economicanalysis.
Implantable pumps will not be covered by thereview. We believe these to be a differenttechnology at an earlier stage of application.Pumps used for external intravenous infusion inhospital care will also be excluded.
It is assumed that the usual sequence is to start onconventional insulin treatment and to move tointensive treatment with MDI if control isinadequate, and so the main question for thisreview is whether it is worth moving fromoptimised MDI to CSII.
We will define intensive or MDI as a combinationof short-acting insulins (soluble or analogue) tocover meals with long-acting insulins(intermediate ultralente or long-acting analoguessuch as glargine) to provide basal insulin. Inpractice, MDI will mean a minimum of threeinjections per day.
Report methodsSearch strategyIn order to capture not only RCTs for efficacyanalysis, but also information on problems with
pumps and reasons for discontinuation, economicstudies, patient experiences and long-termoutcomes, a very sensitive search will be carriedout. Filtering will then be done by reading theabstracts (this has been done for MEDLINE; 1349abstracts have been checked, and about 110studies identified for review of the full paper).Reference lists of retrieved studies will be checkedfor others. Our expert advisers will be asked tocomment on the comprehensiveness of our review.The Cochrane Metabolic and Endocrine DiseasesGroup based in Düsseldorf will be consulted.
Inclusion and exclusion criteriaExclusions will include: conventional therapy;treatment of newly diagnosed patients;implantable pumps; very short-term studies; andhospital in-patient pumps.
Because the key measure of blood glucose controlis glycosylated haemoglobin (HbA1c), studies of<10 weeks duration on each treatment will beexcluded from any HbA1c analysis.
Inclusion and exclusion criteria will be applied byone reviewer and checked by a second. Anydisagreement will be resolved by discussion.
Data extraction strategyData will be extracted by one person and checkedby a second. Any discordances will be resolved bydiscussion or by checking by a third person.
Quality assessment strategyThis will be done in accordance with Chapter II.5of CRD Report 4 (2nd edition). A locallydeveloped system will be used to assess QoLstudies. Criteria will be applied by one reviewerand checked by a second.
Patient perspectivesOur expert panel will include several users of CSII,and we will explore with INPUT how best to includeinformation on patient experiences with CSII.
Methods of analysis/synthesisClinical effectiveness will be synthesised through anarrative review with tabulation of results ofincluded studies.
Combination of studies into a Review Managermeta-analysis will be considered. This could formthe basis for a collaborative Cochrane review.
Economic appraisalThe costs and effects of CSII will be comparedwith MDI in Type 1 diabetes patients. Costs will be
Appendix 1
98
obtained from published literature, NHS sources,ad hoc studies and industry submissions. Costs tobe considered will include NHS resource usebefore and after CSII (e.g. insulin pumps, pumpmaintenance, disposables, extra hospital contacts);educational needs (at both institutional andpatient level); and insulin requirements. If there isevidence of differences in long-term outcomessuch as eye or kidney disease, the economicconsequences of these outcomes will beconsidered. The likely numbers of patients whomight be treated with CSII will be considered. Theperspective of the economic analysis will be that ofthe NHS and Personal and Social Servicesdecision-maker.
Data on clinical and QoL benefits will be soughtfrom the literature. Cost-effectiveness analysis willcompare CSII and MDI on the basis of theprimary outcome measures specified as part of theliterature review (e.g. HbA1c, severe hypoglycaemicevents and diabetes-related complications) andadditional QoL outcomes where documented aspart of the review findings. Information from thepatient impact assessment will also be considered.
Economic analysis will consider� Short-term benefits such as reduction in
hypoglycaemic events, greater flexibility oflifestyle and quality of life
� Short-term disbenefits such as acute events dueto pump failure, cosmetic problems,interference with leisure activities
� Intermediate benefits such as improved blood glucose control as measured by HbA1c,likely to lead to a reduction in long-termcomplications.
If evidence showed that CSII was cost-effective onshort-term outcomes alone, the benefits of longerterm outcomes would not need to be preciselyquantified and they would simply be listed asadditional benefits.
Cost per QALY values will be estimated if the dataprovide evidence of cost-effectiveness gains.
Published cost-effectiveness studies will bereviewed. All papers that present findings on thecost-effectiveness of CSII when compared withMDI (as defined above) will be reviewed in detail,comprising a narrative review with tabulation ofresults where appropriate.
Company submission(s)We will use company and other submissions tocheck on completeness of ascertainment of relevanttrials, for costs of pumps and CSII and for data oncurrent use of insulin pumps in England andWales. We will compare results of cost-effectivenessanalysis from industry models with theSouthampton Health Technology AssessmentCentre one, but in line with our contract the timespent on industry models may be limited to 5person days. This may not allow sufficient time fora detailed critique of industry models.
Health Technology Assessment 2004; Vol. 8: No. 43
99
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
The databases were searched for publishedstudies and recently completed and ongoing
research. All searches were limited to Englishlanguage only.
Clinical effectiveness searchstrategiesCochrane Library (Issue 2, 2002) and NationalResearch Register (Issue 2, 2002):#1 (INSULIN near PUMP*)#2 CSII#3 ((CONTINUOUS near INSULIN) nearINFUSION)#4 ((SUBCUTANEOUS near INSULIN) nearINFUSION)#5 ((EXTERNAL and PUMP*) AND (DIABET*
near INSULIN*))#6 ((((#1 or #2) or #3) or #4) or #5)
MEDLINE (WebSPIRS), 1985–June 2002:((explode ‘Diabetes-Mellitus-Insulin-Dependent’ /all subheadings in MIME,MJME) and ((insulinnear pump*) or (csii) or (continuous near insulinnear infusion) or (subcutaneous near insulin nearinfusion) or (external pump* and diabet* andinsulin*))) and (English in la)
Embase (WebSPIRS), 1980–May 2002:(((insulin near3 pump*) or ((pump* near therapy)and diabet*) or (csii) or (subcutaneous near insulinnear infusion) or (external pump near insulin) or(continuous near insulin near infusion) or (externalpump near diabet*)) and (English in la)) or((external pumps near insulin) and (English in la))
PubMED (Internet version), records added from20 June 2001 to 25 June 2002:insulin pumps* OR CSII OR continuous insulininfusion OR subcutaneous insulin infusion
Science Citation Index, 1990–26 June 2002:insulin pump* or csii or insulin infusion (restrictedto document type = meeting abstracts)
BIOSIS, 1999–26 June 2002:((insulin and pump*) or csii or continuous
subcutaneous insulin)) and random* (restricted todocument type = meeting abstracts)
Web of Science Proceedings,1990–26 June 2002:(insulin pump* or csii or insulin infusion) andrandom*
DARE and HTA Databases (web version), searchedon 29 June 2002:csii or insulin pump$ or insulin infusion
Libcat (in-house library catalogue):insulin pump* or csii or insulin infusion
A flowchart of identification of studies forinclusion in the review of clinical effectiveness isgiven in Figure 17.
Cost effectiveness and quality oflife searchesMEDLINE (WebSPIRS), 1981–June 2002:((insulin near3 pump*) or csii or (insulin nearinfusion)) and ((cost* or economic*) or (explode‘Economics-’ / all subheadings in MIME,MJME) or(explode ‘Health-Status’ / all subheadings inMIME,MJME) or (explode ‘Outcome-Assessment-Health-Care’ / all subheadings in MIME,MJME) or(explode ‘Quality-of-Life’ / all subheadings inMIME,MJME) or (wellbeing or well-being) or(health near3 outcome*))
EMBASE (WebSPIRS), 1980–May 2002:(((insulin near pump*) or (csii or (insulin nearinfusion))) and (((explode ‘health-economics’ / allsubheadings) or (explode ‘economics-’ / allsubheadings) or (explode ‘quality-of-life’ / allsubheadings) or (quality near3 life) or (cost* oreconomic*) or (health near3 status) or (healthnear3 outcome*)) or (wellbeing or well-being)))and (English in la)
PubMED (Internet version, records added from20 June 2001 to 29 June 2002:insulin pumps* OR CSII OR continuous insulininfusion OR subcutaneous insulin infusion
Health Technology Assessment 2004; Vol. 8: No. 43
101
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Appendix 2
Sources of information, including databases searchedand search terms
PsycINFO (WebSPIRS), 1984–June 2002:(insulin near pump*) or (csii or (insulin nearinfusion))
CINAHL (WebSPIRS), 1982–December 2001:((insulin near pump*) or csii or (insulin nearinfusion)) and ((cost* or economic*) or (explode‘Economics-’ / all topical subheadings / all agesubheadings in DE) or (explode ‘Health-Status’ /all topical subheadings / all age subheadings inDE) or (explode ‘Quality-of-Life’ / all topicalsubheadings / all age subheadings in DE) or(wellbeing or well-being) or (explode ‘Outcomes-Health-Care’ / all topical subheadings / all agesubheadings in DE) or (health near3 outcome*))
NHS EED (web version), searched on 29 June2002:csii or insulin pump$ or insulin infusion
EconLit (WebSPIRS): 1969–May 2002csii or insulin pump* or insulin infusion
Additional searchingBibliographies: all references of articles for whichfull papers were retrieved were checked to ensurethat no eligible studies had been missed.
Industry submissions to NICE were examined forany further studies that met the inclusion criteria.
Appendix 2
102
Total identified on searchingn = 3760
Abstracts inspected
Full copies retrievedn = 106
Papers inspected
Papers for appraisal and data extractionCSII vs MDI: n = 22 (20 trials)
Analogue vs soluble: n = 6
Excludedn = 78
Excludedn = 3654
FIGURE 17 Flowchart of identification of studies for inclusion in the review of clinical effectiveness
This is adapted from CRD Report 4 (2nd edition).
Quality criteria for assessment of experimental studies
Health Technology Assessment 2004; Vol. 8: No. 43
103
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Appendix 3
Quality assessment scale
1. Was the assignment to the treatment groups really random?2. Was the treatment allocation concealed?3. Were the groups similar at baseline in terms of prognostic factors?4. Were the eligibility criteria specified?5. Were the point estimates and measure of variability presented for the primary outcome measure?6. Did the analyses include an intention-to-treat analysis?7. Were withdrawals and dropouts completely described?
Some instructions for using a checklist for RCTs
Quality item Coding Explanation
1. Was the assignment to the treatment groups really random?
Random sequence generation Adequate Adequate: random numbers table or computer Partial and central office or coded packagesInadequate Partial: (sealed) envelopes without further Unknown description or serially numbered opaque, sealed
envelopes Inadequate: alternation, case record number, birthdate or similar proceduresUnknown: just the term ‘randomised’ or ‘randomlyallocated’ etc.
2. Was the treatment allocation concealed?
Concealment of randomisation Adequate Adequate: when a paper convinces you that The person(s) who decide on eligibility should Inadequate allocation cannot be predicted [separate persons, not be able to know or be able to predict with Unknown placebo really indistinguishable, clever use of block reasonable accuracy to which treatment group sizes (large or variable)]. Adequate approaches might a patient will be allocated. In trials that use include centralised or pharmacy-controlled good placebos this should normally be the case; randomisation, serially numbered identical however, different modes or timing of drug containers, on-site computer-based system with a administration in combination with the use of randomisation sequence that is not readable until small block sizes of known size may present allocation and other approaches with robust opportunities for clinicians who are also involved methods to prevent foreknowledge of the allocation in the inclusion procedure to make accurate sequence to clinicians and patientsguesses and selectively exclude eligible patients Inadequate: this option is often difficult. You have to in the light of their most likely treatment visualise the procedure and think how people might allocation; in centres with very low inclusion be able to circumvent it. Inadequate approaches frequencies combined with very brief follow-up might include use of alternation, case record times this may also present a potential problem numbers, birth dates or week days, open random because the outcome of the previous patient numbers lists, serially numbered envelopes (even may serve as a predictor of the next likely sealed opaque envelopes can be subject to allocation manipulation) and any other measures that cannot
prevent foreknowledge of group allocationUnknown: no details in text. Disagreements or lackof clarity should be discussed in the review team
continued
Appendix 3
104
Quality item Coding Explanation
3. Were the groups similar at baseline regarding the prognostic factors?
Baseline characteristics Reported Consult the list of prognostic factors or baseline Main aim is to enable the reviewer to see Unknown characteristics (not included in this Appendix).which patients were actually recruited. It Reviewer decidesenables one to get a rough idea on prognostic comparability. A real check on comparability requires multivariable stratification (seldom shown)
4. Were the eligibility criteria specified?
AdequatePartialInadequateUnknown
5. Were the point estimates and measure of variability presented for the primary outcome measure?
Results for the primary outcome measure Adequate Adequate: mean outcome in each group together Partial with mean difference and its standard error (SE) or Inadequate standard deviation (SD) or any CI around it or the Unknown possibility to calculate those from the paper. Survival
curve with log-rank test and patient numbers at latertime pointsPartial: partially reportedInadequate: no SE or SD, or SD without N (SE =SD/N)Unknown: very unlikely
6. Did the analysis include an intention to treat analysis?
Intention-to-treat analysis (ITT) Adequate Reviewers should not just look for the term ITT but Early dropout can make this very difficult. Inadequate assure themselves that the calculations were Strictest requirement is sensitivity analysis according to the ITT principleincluding early dropouts
7. Loss to follow-up
This item examines both numbers and reasons; Adequate Adequate: number randomised must be stated. typically an item that needs checking in the Partial Number(s) lost to follow-up (dropped out) stated or methods section and the marginal totals in the Inadequate deducible (from tables) for each group and reasons tables. Note that it may differ for different Unknown summarised for each groupoutcome phenomena or time points. Some Partial: numbers, but not the reasons (or vice versa) reasons may be reasons given by the patient Inadequate: numbers randomised not stated or not when asked and may not be the true reason. specified for each groupThere is no satisfactory solution for this Unknown: no details in text
Arias P, Kerner W, Zier H, Navascues I, Pfeiffer EF.Incidence of hypoglycemic episodes in diabetic patientsunder continuous subcutaneous insulin infusion andintensified conventional insulin treatment: assessmentby means of semiambulatory 24-hour continuous bloodglucose monitoring. Diabetes Care 1985;8:134–40. [Studydesign]
Barbosa J, Menth L, Eaton J, Sutherland D, Freier EF,Najarian J. Long-term, ambulatory, subcutaneousinsulin infusion versus multiple daily injections in brittlediabetic patients. Diabetes Care 1981;4:269–74.[Insufficient follow-up]
Beck-Nielsen H, Richelsen B, Schwartz-Sorensen N,Hother-Nielsen O. Insulin pump treatment: effect onglucose homeostasis, metabolites, hormones, insulinantibodies and quality of life. Diabetes Res 1985;2:37–43.[Not intensive therapy]
Bell DS, Ovalle F. Improved glycemic control with use ofcontinuous subcutaneous insulin infusion compared withmultiple insulin injection therapy. Endocr Pract 2000;6:357–60. [Study design]
Blackett PR. Insulin pump treatment for recurrentketoacidosis in adolescence. Diabetes Care 1995;18:881–2. [Study design]
Bode BW, Strange P. Efficacy, safety, and pumpcompatibility of insulin aspart used in continuoussubcutaneous insulin infusion therapy in patients withtype 1 diabetes. Diabetes Care 2001;24:69–72.[Insufficient follow-up]
Boland EA, Grey M, Oesterle A, Fredrickson L,Tamborlane WV. Continuous subcutaneous insulininfusion. A new way to lower risk of severehypoglycemia, improve metabolic control, and enhancecoping in adolescents with type 1 diabetes. Diabetes Care1999;22:1779–84. [Study design]
De Beaufort CE, Houtzagers CM, Bruining GJ, Aarsen RS, den Boer NC, Grose WF, et al. Continuoussubcutaneous insulin infusion (CSII) versus conventionalinjection therapy in newly diagnosed diabetic children:two-year follow-up of a randomized, prospective trial.Diabet Med 1989;6:766–71. [Not intensive therapy]
Garg SK, Anderson JH, Gerard LA, Mackenzie TA,Gottlieb PA, Jennings MK, et al. Impact of insulin lisproon HbA1c values in insulin pump users. Diabetes ObesityMetab 2000;2:307–11. [Study design]
Helve E, Koivisto VA, Lehtonen A, Pelkonen R,Huttunen JK, Nikkila EA. A crossover comparison ofcontinuous insulin infusion and conventional injectiontreatment of type I diabetes. Acta Med Scand 1987;221:385–93. [Not intensive therapy]
Jennings AM, Lewis KS, Murdoch S, Talbot JF, Bradley C, Ward JD. Randomized trial comparingcontinuous subcutaneous insulin infusion andconventional insulin therapy in type II diabetic patientspoorly controlled with sulfonylureas. Diabetes Care1991;14:738–44. [Not intensive therapy]
Johansson UB, Adamson UC, Lins PE, Wredling RA.Improved blood glucose variability, HbA1c insumanInfusat and less insulin requirement in IDDM patientsusing insulin lispro in CSII. The Swedish MulticenterLispro Insulin Study. Diabetes Metab 2000;26:192–6.[Insufficient follow-up]
Laatikainen L, Teramo K, Hieta-Heikurainen H,Koivisto V, Pelkonen R. A controlled study of theinfluence of continuous subcutaneous insulin infusiontreatment on diabetic retinopathy during pregnancy.Acta Med Scand 1987;221:367–76. [Not intensivetherapy]
Lauritzen T, Frost LK, Larsen HW, Deckert T. Effect of 1year of near-normal blood glucose levels on retinopathyin insulin-dependent diabetics. Lancet 1983;i:200–4.[Not intensive therapy]
Lepore M, Pampanelli S, Fanelli C, Porcellati F, Bartocci L, Di Vincenzo A, et al. Pharmacokinetics andpharmacodynamics of subcutaneous injection of long-acting human insulin analog glargine, NPH insulin, andultralente human insulin and continuous subcutaneousinfusion of insulin lispro. Diabetes 2000;49:2142–8.[Insufficient follow-up]
Mancuso S, Caruso A, Lanzone A, Bianchi V, Massidda M, Codipietro F, et al. Continuoussubcutaneous insulin infusion (CSII) in pregnantdiabetic women. Acta Endocrinol Suppl (Copenh) 1986;277:112–16. [Study design]
Marshall SM, Home PD, Taylor R, Alberti KG.Continuous subcutaneous insulin infusion versusinjection therapy: a randomised cross-over trial underusual diabetic clinic conditions. Diabet Med1987;4:521–5. [Not intensive therapy]
Ng-Tang FS, Pickup JC, Bending JJ, Collins AC, Keen H, Dalton N. Hypoglycemia andcounterregulation in insulin-dependent diabeticpatients: a comparison of continuous subcutaneousinsulin infusion and conventional insulin injectiontherapy. Diabetes Care 1986;9:221–7. [Not intensivetherapy]
Olsen T, Ehlers N, Nielsen CB, Beck-Nielsen H.Diabetic retinopathy after one year of improvedmetabolic control obtained by continuous subcutaneousinsulin infusion (CSII). Acta Ophthalmol (Copenh)1985;63:315–19. [Not intensive therapy]
Health Technology Assessment 2004; Vol. 8: No. 43
105
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Appendix 4
List of excluded studies
Olsen T, Richelsen B, Ehlers N, Beck-Nielsen H.Diabetic retinopathy after 3 years’ treatment withcontinuous subcutaneous insulin infusion (CSII). ActaOphthalmol (Copenh) 1987;65:185–9. [Not intensivetherapy]
Ooi C, Mullen P, Williams G. Insulin lispro: the idealpump insulin for patients with severe hypoglycemicunawareness? Diabetes Care 1999;22:1598–9. [Studydesign]
Reeves ML, Seigler DE, Ryan EA, Skyler JS. Glycemiccontrol in insulin-dependent diabetes mellitus.Comparison of outpatient intensified conventionaltherapy with continuous subcutaneous insulin infusion.Am J Med 1982;72:673–80. [Insufficient follow-up]
Rizza RA, Gerich JE, Haymond MW, Westland RE, Hall LD, Clemens AH, et al. Control of blood sugar ininsulin-dependent diabetes: comparison of an artificialendocrine pancreas, continuous subcutaneous insulininfusion, and intensified conventional insulin therapy. N Engl J Med 1980;303:1313–8. [Insufficient follow-up]
The Diabetes Control and Complications Trial ResearchGroup. The effect of intensive treatment of diabetes onthe development and progression of long-termcomplications in insulin-dependent diabetes mellitus. N Engl J Med 1993;329:977–86. [Study design]
Appendix 4
106
Bode B, McCulloch K, Strange P. Insulin aspart efficacyand safety compared to buffered regular insulin(Velosulin®) for continuous subcutaneous insulininfusion. Diabetes 2000;49:394.
Chase HP, Saib SZ, MacKenzie T, Hansen M, Garg SK.Postprandial glucose excursions following four methodsof bolus insulin administration using an insulin pump.Diabetes 2001;50:A92.
Conway RJ, Milliken JA. Option or essential tool?Insulin pump therapy in young women. Diabetes2002;51 (Suppl 2):1921.
Devries JH, Snoek FJ, Kostense PJ, Masurel N, HeineRJ. CSII improves glycaemic control and quality of lifein type 1 diabetic patients with long standing poorglycaemic control. A randomized trial. Diabetes 2002;51 (Suppl 2):520.
Galatzer A, Weintrob N, Cohen D, Benzaquen H, RimerA, Ofan R, et al. Treatment satisfaction of type 1 diabeticpatients: CSII v MDI. Diabetes 2002;51 (Suppl 2):2570.
Laffel L, Loughlin C, Ramchandani N, Butler D, Laffel N, Levine BS, et al. Glycemic challenges of pumptherapy (CSII) in youth with type 1 diabetes (T1DM).Diabetes 2001;50:A66–7.
Lenhard MJ, Maser RE. Continuous subcutaneousinsulin infusion (CSII) in patients with type 2 diabetes.Diabetes 2001;50:A440–1.
Lepore M, Pampanelli S, Fanelli CG, Porcellati F,Brunetti P, Bolli GB. Pharmacokinetics and dynamics ofs.c. injection of insulin glargine, NPH and ultralente inT1DM: comparison with CSII. Diabetes 2000;49:36.
Linkeschova R, Raoul M, Bott U, Berger M, Spraul M.Better diabetes control, quality of life and less severehypoglycaemias with insulin pump treatment.Diabetologia 2000;43:748.
Litton J, Rice A, Friedman N, Oden J, Lee M, FreemarkM. Insulin pump therapy in infants and preschoolchildren with type 1 diabetes. Paediatr Res 2002;51(Suppl 1):748.
Mack-Fogg J, Bates S, Faro B, Sciera M, Ippolito K,Orlowski CC, et al. Safe and effective use of CSII in
young children with type 1 diabetes mellitus: PaediatrRes 2002;51 (Suppl 1):712.
Raskin P, Bode B, Marks JB, Hirsch IB, Weinstein R,McGill J, et al. Insulin aspart (IAsp) is as effective incontinuous subcutaneous insulin infusion as in multipledaily injections for patients with type 2 diabetes. Diabetes2001;50:A128.
Reid SM, Lawson ML. Comparison of CSII versusconventional treatment of diabetes with respect tometabolic control. Quality of life and treatmentsatisfaction. Paediatr Res 2002;51 (Suppl 1):713.
Rudolph JW, Hirsck IB. An assessment of continuoussubcutaneous insulin infusion therapy in an academicdiabetes clinic. Diabetes 2000;49:501.
Tubiana-Rufi N, Coutand R, Bloch J, Munz-Licha G,Delcroix C, Limal JM, et al. Efficacy and tolerance ofinsulin lispro in young diabetic children treated withCSII. Diabetologia 2000;43:762.
Turrentine LA, Wallander JL, Roth DL, Bode BW,Trippe BS, Bell DS. The influence on quality of life ofcontinuous subcutaneous insulin infusion (CSII) therapy.Diabetes 2001;50:A49.
Wainstein J, Cohen Y, Gilad R, Raz I, Wexler ID.Treatment of severe insulin resistance in type 2 diabeticswith insulin pumps. Diabetologia 2000;43:789.
Wainstein J, Metzger M, Menuchin O, Cohen Y, Yaffe A,Ravid M, et al. Insulin pump therapy versus multipledaily injections in obese type 2 diabetic patients.Diabetologia 2001;44:94.
Weintrob N, Benzaquen H, Shalitin S, Fayman G,Galatzer A, Dickerman Z, et al. Continuoussubcutaneous insulin infusion versus multiple dailyinjections in children with type 1 diabetes. Diabetologia2001;44:95.
White NH, Hollander AS, Sadler M, Daniels L. Risks and benefits of continuous subcutaneous insulininfusion (CSII) therapy in children. Diabetes 2001;50:A66.
Zloczower M, Cohen J, Zaidise I, Kanter Y. Insulin pump in pregnancy. Diabetes 2000;49:1885.
Health Technology Assessment 2004; Vol. 8: No. 43
107
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Appendix 5
List of recent abstracts
Health Technology Assessment 2004; Vol. 8: No. 43
109
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Appendix 6
Summary of methodology: adults with Type 1 diabetes
Appendix 6
110
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Bak
et a
l.,19
8787
Den
mar
k,sin
gle-
cent
re,
rand
omise
dcr
osso
ver
Dur
atio
n: C
SII
6m
onth
s, M
DI
6m
onth
s
Trea
tmen
t 1:
CSI
I+
sol
uble
.G
rase
by M
S36
Trea
tmen
t 2:
MD
I+
reg
ular
(3 ×
solu
ble,
1 ×
NPH
).Pe
n in
ject
or(N
ovoP
en)
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
of o
utco
mes
:hy
pos
estim
ated
by
the
M-
valu
e of
Sch
lictk
rull,
and
also
freq
uenc
y of
BG
valu
es <
4m
mol
/l –
both
calc
ulat
ed fr
om h
ome-
reco
rded
BG
pro
files
durin
g la
st 3
mon
ths
onea
ch t
reat
men
t re
gim
en.
Patie
nt p
refe
renc
e by
ques
tionn
aire
at
end
ofst
udy
Allo
catio
n to
tre
atm
ent
grou
ps: r
ando
mise
d, h
alf t
o ea
ch t
reat
men
t.N
o fu
rthe
r de
tails
Run-
in/w
ash-
out
perio
d: t
rans
ition
to
CSI
I or
MD
I mad
e du
ring
hosp
italis
atio
n. N
o ru
n-in
or
was
h-ou
t pe
riod
spec
ified
C
ompa
rabi
lity
of t
reat
men
t gr
oups
: not
rep
orte
dA
naly
sis: W
ilcox
on’s
tes
t fo
r pa
ired
diffe
renc
es (2
-tai
led)
. Mea
ns (S
D)
pres
ente
dSa
mpl
e siz
e/po
wer
cal
cula
tion:
not
rep
orte
dG
ener
alisa
bilit
y: m
ainl
y yo
ung
mal
e ad
ults
with
Typ
e 1
Out
com
e m
easu
res:
HbA
1no
t re
port
ed. D
urat
ion
of d
iabe
tes
repo
rted
in m
onth
s. M
ean
gluc
ose
valu
es d
urin
g la
st 3
mon
ths
ofea
ch t
reat
men
t re
gim
en r
epor
ted.
Ave
rage
dai
ly in
sulin
req
uire
men
tat
opt
imum
insu
lin d
osag
e re
port
edRe
ason
s fo
r w
ithdr
awal
s: in
com
patib
ility
with
pum
p re
gim
en (4
CSI
I)Pa
tient
pre
fere
nce:
% w
antin
g to
con
tinue
with
tre
atm
ent
pres
ente
d, n
umer
ator
and
den
omin
ator
not
rep
orte
dC
onfli
ct o
f int
eres
t: N
OVO
Indu
stri,
Den
mar
k pr
ovid
ed N
ovoP
enin
ject
ors
and
infu
sers
, and
per
form
ed la
bora
tory
ana
lyse
s. S
uppo
rted
by A
arus
Uni
vers
ity, D
anish
Dia
betic
Ass
ocia
tion
and
Dan
ish M
edic
alRe
sear
ch C
ounc
il
Popu
latio
n: y
oung
adu
lts; I
DD
M w
ithdu
ratio
n of
>2.
5 y,
dia
gnos
ed b
efor
e th
eag
e of
30
y.C
riter
ia: n
o sig
n of
neu
ropa
thy,
hype
rten
sion
or a
dvan
ced
retin
opat
hy,
abse
nce
of a
lbum
inur
ia, d
iast
olic
blo
odpr
essu
re <
95m
mH
g, a
bsen
ce o
f cat
arac
t,re
tinal
pro
lifer
atio
ns, h
aem
orrh
ages
or
exud
ates
, <5
mic
ro-a
neur
ysm
s in
the
retin
a. C
-pep
tide
–ve
Part
icip
ants
: N
: 20
End
of s
tudy
: 16
Fem
ale:
6.2
5%M
ean
age
(±SD
): 24
(±2)
yD
iabe
tes
dura
tion
mea
n (±
SD):
5.8
(±3.
8) y
Base
line
valu
es: m
ean
(±SD
):In
sulin
: 46
(±16
)IU
/day
Wei
ght:
74.4
(±9.
7)kg
Prev
ious
reg
imen
: pre
viou
sly t
reat
edco
nven
tiona
lly w
ith in
term
edia
te in
sulin
alon
e or
in c
ombi
natio
n w
ith s
hort
-act
ing
insu
lin o
nce
or t
wic
e da
ily
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
111
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Bode
et
al.,
1996
81
USA
, sin
gle-
cent
re, n
on-
rand
omise
dcr
osso
ver
Dur
atio
n: C
SII
mea
n 37
mon
ths,
MD
I min
. of
12m
onth
s
Trea
tmen
t 1:
CSI
I+
sol
uble
Min
imed
506
/504
Trea
tmen
t 2:
MD
I+
reg
ular
Reas
on fo
r C
SII:
(1) s
ubop
timal
glyc
aem
ic c
ontr
olas
sho
wn
byflu
ctua
ting
BGle
vels
thou
ght
tobe
ass
ocia
ted
with
poor
res
pons
e to
inte
rmed
iate
actin
g in
sulin
; (2
) HbA
1c>
8.0%
;(3
) a h
istor
y of
recu
rren
t se
vere
hypo
glyc
aem
ia(d
efin
ed a
s≥
1ep
isode
s/y
ofhy
pos
requ
iring
assis
tanc
e); o
r(4
)hyp
ogly
caem
icun
awar
enes
s
Ass
essm
ent
of o
utco
mes
:w
eigh
t an
d H
bA1c
reco
rded
qua
rter
ly a
t ea
chvi
sit. B
G r
ecor
ded
whe
nse
vere
hyp
o ex
perie
nced
.H
ypos
/dai
ly in
sulin
/sel
f-m
onito
ring
freq
uenc
y/D
KA
= o
btai
ned
at e
ach
visit
from
pat
ient
s’ p
erso
nal l
ogsh
eet
Allo
catio
n to
tre
atm
ent
grou
ps: p
atie
nts
not
rand
omise
dRu
n-in
/was
h-ou
t pe
riod:
all
star
ted
on C
SII a
s in
-pat
ient
s on
dia
bete
sun
it; n
o w
asho
ut p
erio
dC
ompa
rabi
lity
of t
reat
men
t gr
oups
: not
rep
orte
dA
naly
sis: b
asel
ine
valu
es =
last
HbA
1c, b
ody
wei
ght
and
tota
l dai
lyin
sulin
dos
e re
cord
ed w
hen
patie
nts
on M
DI b
efor
e cr
ossin
g ov
er t
oC
SII (
at le
ast
12m
onth
s). S
ingl
e-sa
mpl
e St
uden
t’s t
-tes
t to
eva
luat
ech
ange
from
bas
elin
e. S
ign
test
to
eval
uate
cha
nges
in e
vent
rat
es o
fhy
pos
and
DK
A fr
om t
he M
DI t
o th
e C
SII p
erio
dSa
mpl
e siz
e/po
wer
cal
cula
tion:
not
rep
orte
dG
ener
alisa
bilit
y: a
dults
pre
viou
sly o
n M
DI c
hang
ed t
o C
SII.
Hist
ory
ofse
vere
hyp
os a
nd/o
r hy
po u
naw
aren
ess
Out
com
e m
easu
res:
HbA
1c, b
ody
wei
ght
and
tota
l ins
ulin
dos
eva
lues
. Num
ber
of s
ever
e hy
pos
and
DK
A e
piso
des
durin
g M
DI y
ear
and
durin
g ea
ch C
SII y
ear
Con
flict
of i
nter
est:
1 au
thor
has
rec
eive
d ho
nora
ria fo
r sp
eaki
ngen
gage
men
ts fr
om M
inim
ed, I
nc.,
and
has
rece
ived
gra
nt s
uppo
rt fo
rth
e M
inim
ed Im
plan
tabl
e Pu
mp
stud
y
Popu
latio
n: P
atie
nts
curr
ently
on
CSI
IC
riter
ia: f
rom
a p
opul
atio
n of
255
pts
usin
gC
SII:
min
. of 1
2m
onth
s on
inte
nsiv
eth
erap
y w
ith M
DIs
bef
ore
switc
hing
to
CSI
I; an
d a
min
. of 1
2m
onth
s on
CSI
I afte
rcr
osso
ver
Part
icip
ants
: N
: 55
End
of s
tudy
: 55
Fem
ale:
64%
Mea
n ag
e (±
SD):
39.2
(±12
.9) y
Dia
bete
s du
ratio
n m
ean
(±SD
): 22
.2(±
9.7)
y
Prev
ious
reg
imen
: MD
I
cont
inue
d
Appendix 6
112
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Brin
chm
ann-
Han
sen
et a
l.,19
8876
Oth
erpu
blic
atio
ns:
Dah
l-Jo
rgen
sen
et a
l.,19
8877
and
Dah
l-Jo
rgen
sen
et a
l., 1
98678
Nor
way
, mul
ti-ce
ntre
, RC
T(p
aral
lel)
(Stu
dy =
conv
entio
nal v
sM
DI v
s C
SII;
data
extr
acte
d on
MD
I and
CSI
Ion
ly)
Dur
atio
n: C
SII
mea
n 40
mon
ths
(ran
ge 2
4–60
),M
DI m
ean
41m
onth
s (r
ange
21–4
7)
Trea
tmen
t 1:
CSI
I+
sol
uble
.N
ordi
sk In
fuse
r,A
utos
yrin
ge A
S6C
,Eu
lgy,
Gra
seby
MS3
6 an
d N
ipro
.
Trea
tmen
t 2:
MD
I+
reg
ular
(4 ×
solu
ble,
1 ×
long
-ac
ting)
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
of o
utco
me:
subj
ectiv
e hy
pos
reco
rded
by p
atie
nt. H
bA1
dete
rmin
ed e
very
mon
thfo
r fir
st y
ear
and
ever
yse
cond
mon
th t
here
afte
r
Allo
catio
n to
tre
atm
ent
grou
ps: b
lock
ran
dom
isatio
n w
ithco
mpu
teris
ed m
atch
ing
of c
linic
al b
ackg
roun
d ch
arac
teris
tics
Run-
in/w
ash-
out
perio
d: h
ome
gluc
ose
mon
itorin
g du
ring
2m
onth
sbe
fore
ran
dom
isatio
n sig
nific
antly
red
uced
HbA
1(p
< 0
.01)
in a
llgr
oups
.C
ompa
rabi
lity
of t
reat
men
t gr
oups
: HbA
1, b
ody
wei
ght,
insu
linre
quire
men
t, ag
e, s
ex a
nd d
urat
ion
of d
iabe
tes
com
para
ble
Ana
lysis
: 2-s
ided
Wilc
oxon
ran
k-su
m t
est
to c
ompa
re v
alue
s be
twee
ntr
eatm
ent
grou
ps a
nd 2
-sid
ed W
ilcox
on s
igne
d ra
nk t
est
to c
ompa
reva
lues
with
in t
reat
men
t gr
oups
. Lev
el o
f sig
nific
ance
= 5
%. H
bA1
data
ext
ract
ed fr
om fi
gure
by
revi
ewer
(exc
ept
2-y
data
). SE
Mpr
esen
ted,
SD
cal
cula
ted
by r
evie
wer
. Dat
a at
2y
extr
acte
d fr
omD
ahl-
Jorg
ense
n, 1
986
(tab
le).
Mul
ticen
tre
varia
bilit
y no
t as
sess
edSa
mpl
e siz
e/po
wer
cal
cula
tion:
not
rep
orte
dG
ener
alisa
bilit
y: a
dult
insu
lin d
epen
dent
dia
betic
sO
ther
: fre
quen
cy o
f hom
e BG
mon
itorin
g w
as r
educ
ed a
fter
1st
12m
onth
s an
d ca
used
a t
rend
for
incr
ease
in H
bA1
in a
ll gr
oups
Reas
ons
for
with
draw
als:
sev
ere
hypo
glyc
aem
ia (1
MD
I)C
onfli
ct o
f int
eres
t: st
udy
supp
orte
d by
gra
nts
from
the
Nor
weg
ian
Cou
ncil
for
Scie
nce
and
Hum
aniti
es, O
slo; t
he N
orw
egia
n D
iabe
tes
Ass
ocia
tion,
Oslo
; the
Nor
weg
ian
Cou
ncil
on C
ardi
ovas
cula
rdi
seas
es, O
slo; t
he U
nive
rsity
of O
slo; t
he A
nder
s Ja
hres
Med
ical
Foun
datio
n, O
slo; a
nd N
ordi
sk G
ento
fte A
/S
Popu
latio
n: in
sulin
-dep
ende
nt d
iabe
ticpa
tient
sIn
clus
ion
crite
ria: a
ge 1
8–45
y, d
iabe
tes
befo
re 3
0y
old;
dur
atio
n of
dise
ase
>6
and
<30
y; n
o cl
inic
al s
igns
of n
ephr
opat
hyor
sys
tem
ic h
yper
tens
ion;
no
hist
ory
ofne
urop
athy
; tes
ted
nega
tive
for
C-p
eptid
e.Pt
s w
ith p
rolif
erat
ive
retin
opat
hy w
ere
not
incl
uded
CSI
I par
ticip
ants
: N
: 15
End
of s
tudy
: 15
Fem
ale:
53%
Mea
n ag
e (m
in.–
max
.): 2
6 (1
8–38
) yD
iabe
tes
dura
tion
mea
n (m
in.–
max
.): 1
2.8
(6.4
–23.
3) y
Base
line
valu
es: m
ean
(±SD
):H
bA1:
10.
1 (±
1.55
)% ,
rang
e 8.
0–12
.9In
sulin
: 0.8
(±0.
23)U
/kg
Wei
ght:
68.6
(±9.
3)kg
MD
I par
ticip
ants
:N
: 15
End
of s
tudy
: 14
Fem
ale:
53%
Mea
n ag
e (m
in.–
max
.): 2
6 (1
9-42
) yD
iabe
tes
dura
tion
mea
n (m
in.–
max
.): 1
2.8
(6.8
–20.
8) y
Base
line
valu
es: m
ean
(±SD
):H
bA1:
9.4
(±1.
55)%
, ra
nge
7.0–
11.7
,p
=ns
Insu
lin: 0
.72
(±0.
08)U
/kg
Wei
ght:
71.7
(±10
.1) k
g, p
= n
s
Prev
ious
reg
imen
: tw
o in
ject
ions
of i
nsul
in
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
113
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Chi
asso
n et
al.,
1984
82
Can
ada,
sin
gle-
cent
re, n
on-
rand
omise
dcr
osso
ver
Dur
atio
n: C
SII
3m
onth
s, M
DI
3m
onth
s
Trea
tmen
t 1:
CSI
I+
sol
uble
. Mill
Hill
infu
ser
Trea
tmen
t 2:
CSI
I+
sol
uble
(sol
uble
,pl
us 1
ultr
alen
te).
MD
I adm
inist
ered
by a
utom
atic
jet
inje
ctor
(Med
i-Je
ctor
). Bo
luse
sbe
fore
mea
ls an
dsn
ack
with
>20
gca
rboh
ydra
te
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
of o
utco
mes
:di
abet
ic d
iary
: inc
l. in
sulin
dose
(col
lect
ed b
efor
est
art
of s
tudy
and
at
end
of3-
mon
th p
erio
d).
Gly
cosy
late
d ha
emog
lobi
nm
easu
red
at e
nd o
f eac
h 3-
mon
th p
erio
d
Allo
catio
n to
tre
atm
ent
grou
ps: r
ando
misa
tion
men
tione
d in
abs
trac
tbu
t no
t in
pap
er, w
hich
sta
tes
half
star
ted
with
CSI
I and
hal
f with
MD
I. M
etho
d no
t st
ated
Run-
in/w
ash-
out
perio
d: n
o w
ash-
out
perio
d in
dica
ted.
Pat
ient
s w
ere
hosp
italis
ed (u
nkno
wn
perio
d) w
here
the
y le
arne
d to
ope
rate
the
inte
rven
tion,
mod
ify in
sulin
and
cal
cula
te c
arbo
hydr
ate
cont
ent
Com
para
bilit
y of
tre
atm
ent
grou
ps: i
ndiv
idua
l pat
ient
dat
a pr
esen
ted
for
age,
sex
, dur
atio
n of
dia
bete
s, b
ut t
reat
men
t gr
oup
not
give
nA
naly
sis: s
tatis
tical
com
paris
ons
wer
e m
ade
by m
eans
of p
aire
d t-
test
SEM
pre
sent
ed, S
D c
alcu
late
d by
rev
iew
erSa
mpl
e siz
e/po
wer
cal
cula
tion:
not
rep
orte
dG
ener
alisa
bilit
y: a
dults
with
insu
lin-d
epen
dent
dia
bete
s w
ith s
igns
of
retin
opat
hy, p
revi
ously
on
conv
entio
nal t
hera
pyO
utco
me
mea
sure
s: H
bA1,
dai
ly in
sulin
dos
eO
ther
: num
ber
of in
ject
ions
in M
DI n
ot s
tate
dC
onfli
ct o
f int
eres
t: st
udy
fund
ed in
par
t by
gra
nts
from
Med
ical
Rese
arch
Cou
ncil
(MA
721
1), t
he C
onse
il de
la R
eche
rché
en
Sant
édu
Qué
bec
and
the
NIH
(2R0
1 A
M 2
199-
05)
Popu
latio
n: in
sulin
-dep
ende
nt d
iabe
tics,
all
had
signs
of r
etin
opat
hy, a
ll w
ithin
20%
of
idea
l bod
y w
eigh
tIn
clus
ion
crite
ria: n
ot r
epor
ted
Part
icip
ants
: N
: 12
End
of s
tudy
: 12
Fem
ale:
58.
3%M
ean
age
(min
.–m
ax.):
26.
8 (1
8–40
) yD
iabe
tes
dura
tion
mea
n (m
in.–
max
.): 1
5.3
(4–2
3) y
Base
line
valu
es: m
ean
(±SD
):H
bA1:
11.
9 (±
2.08
)%
Insu
lin: 4
4.7
(±14
.6)I
U/d
Prev
ious
reg
imen
: con
vent
iona
l the
rapy
cont
inue
d
Appendix 6
114
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Haa
kens
et
al.,
1990
83
Nor
way
, sin
gle-
cent
re, n
on-
rand
omise
dcr
osso
ver
(Stu
dy =
conv
entio
nal v
sM
DI v
s C
SII;
data
extr
acte
d on
MD
I and
CSI
Ion
ly)
Dur
atio
n: C
SII
6m
onth
s, M
DI
6m
onth
s
Trea
tmen
t 1:
CSI
I+
sol
uble
.N
ordi
sk in
fuse
r/(n
= 5
0),
auto
syrin
ge A
S8M
P (n
= 2
)
Trea
tmen
t 2:
MD
I+
reg
ular
[4
×so
lubl
e (v
iaN
ovop
en);
1 ×
inte
rmed
iate
(ultr
alen
te a
ndiso
phan
e us
ed v
iasy
ringe
)]
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
of o
utco
mes
: ou
tpat
ient
clin
ic v
isits
=6
wee
ks, o
ccur
renc
e of
mild
and
ser
ious
hyp
os,
DK
A, i
njec
tion/
infu
sion
site
infe
ctio
ns, r
esul
ts o
f hom
eBG
mon
itorin
g re
cord
ed
Allo
catio
n to
tre
atm
ent
grou
ps: i
ndiv
idua
ls no
t ra
ndom
ised
Run-
in/w
ash-
out
perio
d: r
un-in
per
iod
= 3
–9 m
onth
sC
ompa
rabi
lity
of t
reat
men
t gr
oups
: no
base
line
char
acte
ristic
spr
ovid
ed t
o al
low
com
paris
ons.
No
stat
istic
al s
igni
fican
ce b
etw
een
subj
ects
who
ent
ered
the
stu
dy a
nd t
hose
who
did
n’t
with
reg
ard
toH
bA1
(11.
1 ±
0.29
% v
s. 1
0.9
±0.
16%
) or
age
(24.
4 ±
1.0
y vs
25.
5±
0.5
y)A
naly
sis: n
ot IT
T a
naly
sis. R
esul
ts g
iven
as
mea
ns ±
SEM
. Rev
iew
erca
lcul
ated
SD
. 2-s
ided
Wilc
oxon
sig
ned-
rank
tes
t fo
r pa
ired
data
use
dto
com
pare
diff
eren
t tr
eatm
ent
mod
es. S
pear
man
’s r
ank
corr
elat
ion
coef
ficie
nt u
sed.
Sta
tistic
al s
igni
fican
ce s
et t
o p
< 0
.05.
Sam
ple
size/
pow
er c
alcu
latio
n: n
ot r
epor
ted.
Gen
eral
isabi
lity:
adu
lts (1
6–50
y) w
ith in
sulin
-dep
ende
nt d
iabe
tes
for
>1
y, w
ith n
o pr
olife
rativ
e/pr
epro
lifer
ativ
e re
tinop
athy
Out
com
e m
easu
res:
sta
ble
HbA
1=
ref
eren
ce r
ange
5.4
–7.6
%Re
ason
s fo
r w
ithdr
awal
s: s
erio
us h
ypo
(1 M
DI),
una
ble
to c
ontr
ol B
G(2
MD
I), d
iscon
tinue
d af
ter
1.5
mon
ths
– un
com
fort
able
(6 C
SII),
med
ical
rea
sons
(2 C
SII),
ref
used
to
give
up
pum
p to
go
on M
DI (
4),
refu
sed
tria
l of C
SII (
2)C
onfli
ct o
f int
eres
t: N
orw
egia
n D
iabe
tic A
ssoc
iatio
n pr
ovid
edfin
anci
al s
uppo
rt
Popu
latio
n: in
sulin
-dep
ende
nt d
iabe
ticpa
tient
s ag
ed 1
6–50
y, a
tten
ding
out
patie
ntcl
inic
at
hosp
ital
Excl
usio
n cr
iteria
: on
inte
nsifi
ed in
sulin
ther
apy;
pre
gnan
t or
pla
nnin
g pr
egna
ncy;
diab
etes
dur
atio
n <
1y;
tot
al d
aily
insu
lindo
se <
24U
; ser
um c
reat
inin
e>
150
µmol
/l; p
rolif
erat
ive/
prep
rolif
erat
ive
retin
opat
hy; m
acro
vasc
ular
com
plic
atio
ns;
seve
rely
affe
cted
abi
lity
to r
ecog
nise
hypo
glyc
aem
ia, a
s ev
iden
ced
by m
ultip
lehy
pogl
ycae
mic
com
as w
ithou
t w
arni
ngsy
mpt
oms
durin
g th
e la
st fe
w y
ears
;m
alig
nant
dise
ase;
kno
wn
alco
hol/d
rug
abus
e; p
sych
osis;
men
tal r
etar
datio
n
Part
icip
ants
: N
: 52
End
of s
tudy
: 35
Fem
ale:
63%
Mea
n ag
e (±
SD):
24.9
(±6.
78) y
Dia
bete
s du
ratio
n m
ean
(±SD
): 9.
2(±
5.55
) yBa
selin
e va
lues
: mea
n (±
SD):
HbA
1: 1
0.4
(±1.
85)%
.In
sulin
: 0.7
6 (±
0.25
)IU
/day
BMI:
23.2
(±2.
47)k
g/m
2(a
t en
try
to s
tudy
)W
eigh
t: 68
.6 (±
8.63
)kg.
HbA
1at
beg
inni
ng o
f pre
-per
iod
=10
.7(±
2.47
)
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
115
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Han
aire
-Bro
utin
et a
l., 2
00080
Fran
ce, s
ingl
e-ce
ntre
,ra
ndom
ised
cros
sove
r
Dur
atio
n: C
SII
4m
onth
s, M
DI
4m
onth
s
Trea
tmen
t 1:
CSI
I+
lisp
ro. M
inim
ed50
6/50
7,D
isetr
onic
Htr
onD
/V
Trea
tmen
t 2:
MD
I+
lisp
ro [3
×lis
pro,
2 ×
NPH
(mea
n 2.
65in
ject
ions
at
end
of M
DI)]
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
of o
utco
mes
:da
ily B
G m
easu
rem
ents
–in
eve
nt o
f hyp
ogly
caem
icsy
mpt
oms.
Pts
rec
orde
d al
lep
isode
s of
hyp
ogly
caem
iaan
d an
y ot
her
tech
nica
l or
met
abol
ic in
cide
nt. V
isits
ever
y 2
mon
ths,
insu
lindo
se a
nd a
dver
se e
vent
sno
ted.
Pat
ient
pre
fere
nce
at e
nd o
f stu
dy. H
bA1c
dete
rmin
ed a
t en
d of
eac
hpe
riod
Hyp
os (B
G <
60m
g/dl
)du
ring
last
14
days
Allo
catio
n to
tre
atm
ent
grou
ps: r
ando
mise
d af
ter
run-
in p
erio
d,m
etho
d no
t st
ated
. Allo
catio
n co
ncea
lmen
t un
clea
r, op
en-la
bel d
esig
nRu
n-in
/was
h-ou
t pe
riod:
all
pts
had
6w
eeks
run
-in w
ith C
SII +
regu
lar
insu
lin. N
o w
ash-
out
perio
d. S
tate
s th
at n
o ca
rryo
ver
effe
ctw
as o
bser
ved
Com
para
bilit
y of
tre
atm
ent
grou
ps: n
ot r
epor
ted
Ana
lysis
: res
ults
giv
en a
s m
eans
and
SD
. All
test
s w
ere
2-ta
iled,
and
pva
lues
<0.
05 =
sta
tistic
ally
sig
nific
ant.
Cha
nges
in c
ontin
uous
crit
eria
stud
ied
by a
naly
sis o
f var
ianc
e ap
plie
d to
cro
ssov
er s
tudy
with
perio
d, t
reat
men
t gr
oup,
and
inte
ract
ion
fact
ors.
Cat
egor
ical
crit
eria
(adv
erse
eve
nts)
wer
e co
mpa
red
with
χ2
test
.Sa
mpl
e siz
e/po
wer
cal
cula
tion:
not
rep
orte
dG
ener
alisa
bilit
y: a
dults
with
Typ
e I d
iabe
tes
and
expe
rienc
e of
inte
nsifi
ed t
hera
py (C
SII o
r M
DI)
Out
com
e m
easu
res:
HbA
1cat
4 m
onth
s, in
sulin
dos
e, h
ypo
episo
des
durin
g la
st 1
4da
ysRe
ason
s fo
r w
ithdr
awal
s: d
iffic
ultie
s w
ith M
DI (
pts
on C
SII b
efor
een
rolm
ent)
(1 M
DI)
Con
flict
of i
nter
est:
stud
y w
as s
uppo
rted
by
Lilly
Fra
nce
and
Baye
rFr
ance
Popu
latio
n: T
ype
1 di
abet
ic p
atie
nts
aged
21–6
5y.
Incl
usio
n cr
iteria
: HbA
1c<
10.
0%,
C-p
eptid
e –v
e an
d ex
perie
nce
ofin
tens
ified
insu
lin t
hera
py. N
one
had
untr
eate
d re
tinop
athy
, im
paire
d re
nal
func
tion,
gas
tric
neu
ropa
thy,
BMI>
30kg
/m2 , d
aily
insu
lin d
ose
>2
U/k
g, h
istor
y of
hyp
ogly
caem
iaun
awar
enes
s, o
r an
y se
vere
dise
ase
Part
icip
ants
: N
: 41
End
of s
tudy
: 40
Fem
ale:
48.
8%M
ean
age
(±SD
): 43
.5 (±
10.3
)y, m
in. 2
1,m
ax. 6
5.4
yD
iabe
tes
dura
tion
mea
n (±
SD):
20(±
11.3
y), m
in. 4
, max
. 42
yBa
selin
e va
lues
: mea
n (±
SD):
HbA
1c: 8
.39
(±0.
87)%
Insu
lin: 4
3.6
(±13
.5) U
/day
BMI:
24 (±
2.4)
kg/m
2
Wei
ght:
68.2
(±10
)kg
Prev
ious
reg
imen
: at
enro
lmen
t 32
wer
etr
eate
d w
ith C
SII +
reg
ular
insu
lin a
nd 9
by
MD
I with
reg
ular
insu
lin o
r lis
pro
cont
inue
d
Appendix 6
116
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Hom
e et
al.,
1982
88
UK
, sin
gle-
cent
re,
rand
omise
dcr
osso
ver
Dur
atio
n: C
SII
2.5
mon
ths,
MD
I2.
5 m
onth
s
Trea
tmen
t 1:
CSI
I+
sol
uble
. Mill
Hill
1001
HM
,A
utos
yrin
geA
S*6C
Trea
tmen
t 2:
MD
I+
reg
ular
(2 ×
solu
ble,
1 ×
ultr
alen
te)
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
of o
utco
mes
:nu
mbe
r of
hyp
os r
equi
ring
oral
glu
cose
rec
orde
d at
biw
eekl
y cl
inic
visi
tsPa
tient
pre
fere
nce
byqu
estio
nnai
re a
t en
d of
stud
y
Allo
catio
n to
tre
atm
ent
grou
ps: s
tate
s re
gim
ens
in ‘r
ando
m o
rder
’bu
t no
furt
her
deta
ilsRu
n-in
/was
h-ou
t pe
riod:
not
rep
orte
dC
ompa
rabi
lity
of t
reat
men
t gr
oups
: not
rep
orte
dA
naly
sis: n
ot IT
T. D
ata
pres
ente
d on
10
pts
who
com
plet
ed s
tudy
.M
ean
(SE)
pre
sent
ed. S
D c
alcu
late
d by
rev
iew
er. S
tude
nt’s
pai
red
t-te
stSa
mpl
e siz
e/po
wer
cal
cula
tion:
not
rep
orte
dG
ener
alisa
bilit
y: a
dults
with
insu
lin-d
epen
dent
dia
bete
sO
utco
me
mea
sure
s: b
lood
tak
en fo
r H
bA1
biw
eekl
y an
d fo
rch
oles
tero
l at
end
of e
ach
trea
tmen
t. H
ypos
rec
orde
d at
clin
ic v
isit,
ther
efor
e su
bjec
t to
rec
all b
ias.
Que
stio
nnai
re fo
r pa
tient
pre
fere
nce.
Reas
ons
for
with
draw
als:
1 p
t ex
clud
ed b
ecau
se o
f isc
haem
ic h
eart
dise
ase
(not
cle
ar w
heth
er t
his
was
bef
ore
or a
fter
rand
omisa
tion)
;1
pt h
ad m
yoca
rdia
l inf
arct
ion
befo
re c
ompl
etin
g st
udy
Con
flict
of i
nter
est:
pers
onal
sup
port
from
Nov
o La
bora
torie
sLi
mite
d. F
inan
cial
sup
port
from
Brit
ish D
iabe
tic A
ssoc
iatio
n
Popu
latio
n: in
sulin
-dep
ende
nt, C
-pep
tide
–ve,
adu
lts w
hose
con
trol
had
bee
npr
evio
usly
opt
imise
d on
tw
ice-
daily
inje
ctio
n th
erap
y.C
riter
ia: n
ot s
peci
fied.
1 (o
f 12)
pat
ient
excl
uded
due
to
ischa
emic
hea
rt d
iseas
e
Part
icip
ants
: N
: 12
cons
ente
d (s
ee m
etho
dolo
gy)
End
of s
tudy
: 10
Fem
ale:
40%
Mea
n ag
e (m
in.–
max
.): 4
0.4
(29–
52) y
Dia
bete
s du
ratio
n m
ean
(min
.–m
ax.):
22.
3(7
–34)
yBa
selin
e va
lues
: mea
n (±
SD):
HbA
1: 1
0.7
(±1.
9)%
Insu
lin: 6
0 (±
18.9
7) IU
/day
Wei
ght:
67.6
(±8.
5)kg
Prev
ious
reg
imen
: opt
imise
d on
tw
ice-
daily
inje
ctio
ns
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
117
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Nat
han
and
Lou,
1982
89
USA
, sin
gle-
cent
re,
rand
omise
dcr
osso
ver
Dur
atio
n: e
ach
perio
d8–
12w
eeks
dura
tion;
1su
bjec
t on
MD
Ifo
r <
10w
eeks
.
1 su
bjec
tcr
osse
d-ov
er3
times
(M
DI 8
wee
ks,
CSI
I 12
wee
ks,
MD
I 8w
eeks
,C
SII 4
wee
ks
Trea
tmen
t 1:
CSI
I+
sol
uble
.A
utos
yrin
geA
S*6C
Trea
tmen
t 2:
MD
I+
reg
ular
(3 ×
solu
ble,
2 ×
NPH
or 1
×ul
tral
ente
).
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
of o
utco
mes
:di
ary:
insu
lin d
ose
adju
stm
ents
,hy
pogl
ycae
mic
rea
ctio
ns,
unus
ual a
ctiv
ities
,co
mpl
icat
ions
of t
hera
py.
Hyp
ogly
caem
ia:
com
bina
tion
of s
ympt
oms
of p
alpi
tatio
ns, d
iaph
ores
is,hu
nger
, lig
ht-h
eade
dnes
s or
conf
usio
n oc
curr
ing
with
BG <
60m
g/dl
and
relie
ved
by c
arbo
hydr
ate
inge
stio
nA
ccep
tabi
lity
of t
reat
men
tas
sess
ed w
ithqu
estio
nnai
re
Allo
catio
n to
tre
atm
ent
grou
ps: r
ando
mise
d af
ter
base
line
data
obta
ined
. Met
hod
not
stat
ed. C
once
alm
ent
of a
lloca
tion
not
indi
cate
d1
pt p
artic
ipat
ed in
a d
oubl
e-bl
ind
cros
sove
r tr
ial o
f bot
h th
erap
ies
(3w
eeks
per
arm
)Ru
n-in
/was
h-ou
t pe
riod:
No
indi
catio
n of
was
h-ou
t pe
riod.
Pts
adm
itted
to
IP fo
r 2–
3da
ys a
fter
each
tre
atm
ent
star
ted
and
afte
rcr
osso
ver.
Afte
r co
mpl
etin
g fir
st le
g of
tria
l, pt
s ha
d al
l bas
elin
est
udie
s ag
ain,
inpa
tient
mon
itorin
g w
as c
arrie
d ou
t an
d pt
s w
ere
swap
ped
over
to
the
alte
rnat
ive
ther
apy
Com
para
bilit
y of
tre
atm
ent
grou
ps: b
asel
ine
data
tab
le g
iven
, tho
ugh
com
para
bilit
y be
twee
n gr
oups
(tho
se o
n M
DI f
irst
vs C
SII f
irst)
not
give
nA
naly
sis: S
tude
nt’s
pai
red
2-ta
iled
t-te
st. I
ndiv
idua
l pt
data
pre
sent
ed,
mea
ns c
alcu
late
d by
rev
iew
er. B
asel
ine
and
insu
lin a
nd H
bA1c
(for
trea
tmen
ts) S
D c
alcu
late
d by
rev
iew
erSa
mpl
e siz
e/po
wer
cal
cula
tion:
not
rep
orte
dG
ener
alisa
bilit
y: m
otiv
ated
adu
lts w
ith T
ype
I dia
bete
s, p
revi
ously
on
NPH
and
reg
ular
insu
lin r
egim
esO
utco
me
mea
sure
s: a
ppro
pria
te o
utco
me
mea
sure
s ( H
bA1c
,hy
pogl
ycae
mic
epi
sode
s) u
sed
Con
flict
of i
nter
est:
gran
t su
ppor
t: in
par
t by
Cor
ning
Med
ical
and
Scie
ntifi
c, M
edfie
ld, M
A; t
he M
uria
l McL
auth
lin F
und,
Mas
sach
uset
tsG
ener
al H
ospi
tal;
and
the
Gen
eral
Clin
ical
Res
earc
h C
entr
e gr
ant
RR01
066,
Nat
iona
l Ins
titut
es o
f Hea
lth. O
ne a
utho
r is
the
Cap
psSc
hola
r of
Har
vard
Uni
vers
ity; o
ne a
utho
r is
an in
vest
igat
or fo
r th
eH
owar
d H
ughe
s M
edic
al In
stitu
te
Popu
latio
n: m
otiv
ated
adu
lts w
ith T
ype
Idi
abet
esC
riter
ia: p
ts h
ad h
istor
y of
ket
osis-
pron
e,in
sulin
-dep
ende
nt d
iabe
tes.
Non
e ha
dsig
nific
ant
neur
opat
hy n
ephr
opat
hy,
retin
opat
hy o
r va
scul
ar d
iseas
e at
stu
dyou
tset
. All
pts
with
in 1
0% o
f ide
al b
ody
wei
ght
Part
icip
ants
: N
: 5En
d of
stu
dy: 5
Fem
ale:
60%
Mea
n ag
e (m
in.–
max
.): 3
1 (2
7–41
) yD
iabe
tes
dura
tion
mea
n (m
in.–
max
.):7.
4(5
–9) y
Base
line
valu
es: m
ean
(±SD
):H
bA1c
: 9.0
3 (±
1.43
)%.
Insu
lin: 4
2.8
(±16
.04)
IU/d
ayH
ypog
lyca
emic
rea
ctio
ns (n
/wee
k) =
1.8
(±0.
84)
Prev
ious
insu
lin r
egim
en: r
egul
ar +
NPH
(2 ×
daily
) (3
pts)
; NPH
(2 ×
daily
) (1
pt);
regu
lar
+ N
PH (1
pt)
cont
inue
d
Appendix 6
118
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Nos
adin
i et
al.,
1988
85
Italy,
sin
gle
cent
re, R
CT
(par
alle
l)
[4 a
rms:
conv
entio
nal v
sM
DI v
s C
SII w
ithfix
ed b
asal
rat
e(F
BR) v
s C
SII
with
hig
her
daw
nin
sulin
(HO
R).
Dat
a on
conv
entio
nal a
rmno
t ex
trac
ted]
Dur
atio
n:12
mon
ths
Trea
tmen
t 1:
CSI
I+
sol
uble
(FBR
).M
icro
jet
MC
20
Trea
tmen
t 2:
MD
I+
reg
ular
(3 ×
solu
ble,
1 ×
long
-ac
ting)
Trea
tmen
t 3:
CSI
I+
sol
uble
(HO
R).
Beta
tron
II
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
of o
utco
mes
:BG
mea
sure
d w
hen
hype
ran
d hy
po e
piso
des
susp
ecte
d. E
vent
s as
sess
edfr
om t
wic
e-w
eekl
ypr
ofile
s, h
ospi
tal B
G a
nd p
tre
cord
s of
sus
pect
edev
ents
. Hyp
er =
BG
up
to35
0m
g/dl
with
ace
tonu
riaas
sess
ed b
y pt
, no
med
ical
advi
ce. K
etot
ic e
vent
s =
BG >
400
mg/
dl a
ndke
tone
bod
y le
vels
>1.
5m
mol
/l w
ith c
linic
alsy
mpt
oms
requ
iring
adm
issio
n to
war
d. H
ypo
= B
G<
60m
g/dl
with
clin
ical
sym
ptom
s w
ithou
the
lp fr
om o
ther
s. S
ever
ehy
po =
BG
<60
mg/
dlre
quiri
ng g
luca
gon
orgl
ucos
e pa
rent
eral
adm
inist
ratio
n fr
om o
ther
s
Allo
catio
n to
tre
atm
ent
grou
ps: 9
6 pt
s re
crui
ted
and
divi
ded
into
4gr
oups
: mal
e, fe
mal
e, d
urat
ion
of d
iseas
e >
5y,
<5
y. S
ubje
cts
ofea
ch g
roup
wer
e as
signe
d ‘b
lindl
y’ t
o ea
ch o
f 4 t
reat
men
ts. M
etho
dno
t st
ated
. Sea
led
enve
lope
s op
ened
by
pts.
Con
sent
obt
aine
d af
ter
rand
omisa
tion
Run-
in/w
ash-
out
perio
d: a
fter
rand
omisa
tion,
met
abol
ic d
ata
colle
cted
for
3 m
onth
s on
pre
viou
s tr
aditi
onal
the
rapy
. A w
eek
inho
spita
l was
spe
nt t
each
ing
self-
mon
itorin
g an
d se
lf-re
gula
tion
ofin
sulin
Com
para
bilit
y of
tre
atm
ent
grou
ps: s
tate
s sim
ilar
insu
lin d
aily
dos
ean
d nu
mbe
r of
inje
ctio
ns, a
nd H
bA1c
Ana
lysis
: no
cros
sove
r w
ithin
tre
atm
ents
occ
urre
d. M
ultiv
aria
teA
NO
VA fo
r re
peat
ed m
easu
res
used
for
diffe
renc
es w
ithin
the
4in
sulin
reg
imen
s an
d be
twee
n tim
e 0
and
each
reg
imen
. Unp
aire
d t-
test
ana
lysis
and
Bon
ferr
oni t
-tes
t fo
r di
ffs b
etw
een
singl
e va
lues
whe
n A
NO
VA r
evea
led
signi
fican
t di
ffere
nce
Sam
ple
size/
pow
er c
alcu
latio
n: n
ot r
epor
ted
Gen
eral
isabi
lity:
insu
lin-d
epen
dent
dia
betic
s re
ceiv
ing
conv
entio
nal
ther
apy
Out
com
e m
easu
res:
bas
elin
e va
lues
col
lect
ed in
mon
th p
rece
ding
time
0. H
ypos
exp
ress
ed a
s m
ean
valu
e/pt
/yea
rC
onfli
ct o
f int
eres
t: su
ppor
ted
by C
NR
gran
ts a
nd C
NR
Inte
rnat
iona
lRe
sear
ch P
rogr
amm
e
Popu
latio
n: in
sulin
-dep
ende
nt d
iabe
tics
Incl
usio
n cr
iteria
: C-p
eptid
e –v
e
CSI
I FBR
par
ticip
ants
: N
: 19
End
of s
tudy
: 19
Fem
ale:
42.
1%M
ean
age
(±SD
): 36
(±6)
yD
iabe
tes
dura
tion
mea
n (±
SD):
8 (±
3)y
Base
line
valu
es: m
ean
(±SD
):W
eigh
t: 77
(±7)
kg
MD
I par
ticip
ants
:N
: 15
End
of s
tudy
: 15
Fem
ale:
26.
7%M
ean
age
(±SD
): 32
(±9)
yD
iabe
tes
dura
tion
mea
n (±
SD):
7 (±
4)y
Base
line
valu
es: m
ean
(±SD
):W
eigh
t: 71
(±6)
kg
CSI
I HO
R pa
rtic
ipan
ts:
N: 1
0En
d of
stu
dy: 1
0Fe
mal
e: 4
0%M
ean
age
(±SD
): 34
(±3)
yD
iabe
tes
dura
tion
mea
n (±
SD):
7 (±
3)y
Base
line
valu
es: m
ean
(±SD
):W
eigh
t: 70
(±7)
kg
Prev
ious
reg
imen
: 2–3
inje
ctio
ns/d
ay,
mix
ed r
egul
ar a
nd lo
ng-a
ctin
g in
sulin
prep
arat
ions
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
119
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Saur
brey
et
al.,
1988
90
Den
mar
k, s
ingl
e-ce
ntre
,ra
ndom
ised
cros
sove
r
Dur
atio
n: C
SII
2.5
mon
ths,
MD
I2.
5m
onth
s
Trea
tmen
t 1:
CSI
I+
sol
uble
.A
utos
yrin
ge A
S-6C
(n=
6),
Med
ix20
9 (n
= 1
3).
Trea
tmen
t 2:
MD
I+
reg
ular
[3
×so
lubl
e(N
ovop
en),
1 ×
inte
rmed
iate
(syr
inge
)].
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
of o
utco
mes
:tw
ice
wee
kly
= 7
-poi
ntBG
pro
files
. Num
ber
ofhy
pos
reco
rded
. No
furt
her
deta
ils.
Bloo
d sa
mpl
es fo
r H
bA1c
take
n af
ter
run-
in a
nd a
fter
each
tre
atm
ent
perio
dA
t al
l out
patie
nt v
isits
=in
ject
ions
site
s in
spec
ted
and
body
wei
ght
reco
rded
.Pa
tient
att
itude
ques
tionn
aire
com
plet
ed a
ten
d of
stu
dy
Allo
catio
n to
tre
atm
ent
grou
ps: p
ts r
ando
mise
d, b
ut n
o in
dica
tion
ofm
etho
d of
ran
dom
isatio
n. N
o m
entio
n of
con
ceal
men
t of
allo
catio
nRu
n-in
/was
h-ou
t pe
riod:
pts
had
a 3
-wee
k ru
n-in
Com
para
bilit
y of
tre
atm
ent
grou
ps: n
ot r
epor
ted
Ana
lysis
: Mea
n an
d ±
SE v
alue
s pr
ovid
ed. R
evie
wer
cal
cula
ted
SD fo
rH
bA1c
and
insu
lin. R
esul
ts c
ompa
red
stat
istic
ally
by
Stud
ent’s
t-t
est
and
diffe
renc
es r
esul
ting
in p
-val
ues
<0.
05 w
ere
cons
ider
edsig
nific
ant.
HbA
1cda
ta e
xtra
cted
from
figu
re b
y re
view
erSa
mpl
e siz
e/po
wer
cal
cula
tion:
not
rep
orte
dG
ener
alisa
bilit
y: a
dults
with
insu
lin-d
epen
dent
dia
bete
s w
ho h
adpr
evio
usly
bee
n on
inje
ctio
n in
sulin
the
rapy
.O
utco
me
mea
sure
s: a
ppro
pria
te o
utco
me
mea
sure
s us
ed (H
bA1c
and
hypo
s), a
lthou
gh n
o in
dica
tion
of r
ecor
ding
met
hod
of h
ypos
Reas
ons
for
with
draw
als:
pum
p w
earin
g un
acce
ptab
le (2
)C
onfli
ct o
f int
eres
t: N
ovo
Indu
stri
A/S
sup
plie
d N
ovoP
ens
Popu
latio
n: a
dults
with
insu
lin d
epen
dent
diab
etes
; ass
umed
tha
t m
ost
of t
he p
ts C
-pe
ptid
e –v
e. A
ll w
ere
trea
ted
with
tw
ice
daily
inje
ctio
ns o
f sho
rt-a
ctin
g pl
usin
term
edia
te-
or lo
ng-a
ctin
g in
sulin
.Ba
ckgr
ound
ret
inop
athy
(12)
, pro
lifer
ativ
ere
tinop
athy
(0),
inci
pien
t se
nsor
yne
urop
athy
(4),
inte
rmitt
ent
prot
einu
ria (3
)In
clus
ion
crite
ria: i
nsul
in-d
epen
dent
diab
etes
; abl
e to
com
ply
with
inte
nsiv
ein
sulin
reg
ime
Part
icip
ants
: N
: 21
End
of s
tudy
: 19
Fem
ale:
43%
Mea
n ag
e (±
SD):
32 (±
2.1)
y, m
in. 2
0,m
ax. 5
3D
iabe
tes
dura
tion
mea
n (±
SD):
14.5
(±1.
4) y
Base
line
valu
es: m
ean
(±SD
):G
roup
sta
rtin
g w
ith C
SII,
HbA
1c: 8
.7(±
1.74
)%G
roup
sta
rtin
g w
ith M
DI H
bA1c
: 8.8
(±2.
17)%
, p=
ns
Insu
lin: 0
.67
(±0.
4) IU
/day
cont
inue
d
Appendix 6
120
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Schi
ffrin
and
Belm
onte
, 198
291
Can
ada,
sin
gle-
cent
re,
rand
omise
dcr
osso
ver
Dur
atio
n: C
SII
6m
onth
s, M
DI
6m
onth
s
Trea
tmen
t 1:
CSI
I+
sol
uble
. Mill
Hill
1001
GM
Trea
tmen
t 2:
MD
I+
reg
ular
(3 ×
solu
ble,
1 ×
NPH
).
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
of o
utco
mes
:H
ypos
: mild
= r
ecog
nise
dby
sen
satio
n of
hun
ger,
fatig
ue, i
rrita
bilit
y,he
adac
he. M
oder
ate
=co
ld s
wea
ting,
pal
pita
tions
,tr
emou
rs, b
lurr
ed v
ision
,im
paire
d vo
lunt
ary
mov
emen
ts o
r m
emor
ylo
ss. S
ever
e =
loss
of
cons
ciou
snes
s. M
etho
d of
colle
ctin
g pt
pre
fere
nce
data
not
rep
orte
d. H
bA1
and
chol
este
rol m
easu
red
mon
thly
Allo
catio
n to
tre
atm
ent
grou
ps: p
ts r
ando
mise
d, h
alf t
o ea
ch t
hera
py,
but
no in
dica
tion
of m
etho
d of
ran
dom
isatio
n. A
lloca
tion
conc
ealm
ent
uncl
ear
Run-
in/w
ash-
out
perio
d: p
ts r
emai
ned
on c
onve
ntio
nal t
reat
men
tpr
ior
to s
tudy
. Pts
sta
rtin
g on
CSI
I wer
e gi
ven
prac
tice
for
1w
eek
prio
r to
stu
dy. N
o w
ash-
out
perio
d re
port
edC
ompa
rabi
lity
of t
reat
men
t gr
oups
: not
rep
orte
dA
naly
sis: n
ot IT
T a
naly
sis. M
ean
±SD
pro
vide
d. D
ata
anal
ysed
by
paire
d St
uden
t’s t
-tes
t. W
hen
valu
es d
id n
ot h
ave
a no
rmal
dist
ribut
ion,
a p
aire
d t-
test
was
per
form
ed o
n th
e lo
garit
hmic
tran
sfor
m o
f the
dat
a. B
asel
ine
and
6-m
onth
insu
lin v
alue
sex
trap
olat
ed fr
om g
raph
(Fig
ure
4) b
y re
view
er. B
asel
ine
wei
ght
and
BMI c
alcu
late
d fr
om T
able
1 b
y re
view
erSa
mpl
e siz
e/po
wer
cal
cula
tion:
not
rep
orte
dG
ener
alisa
bilit
y: in
sulin
-dep
ende
nt d
iabe
tics
aged
14–
41 y
who
had
prev
ious
ly b
een
on c
onve
ntio
nal i
njec
tion
insu
lin t
hera
pyO
utco
me
mea
sure
s: H
bA1
repo
rted
mon
thly
for
mon
ths
0–6.
Dat
aex
trac
ted
for
mon
ths
0, 3
and
6 o
nly.
Met
hod
of r
ecor
ding
hyp
os n
otcl
ear.
Pts
follo
wed
the
ir us
ual d
iabe
tic d
iet:
30–4
0% fa
t, 15
–20%
prot
ein,
40–
45%
CH
O in
the
form
of 3
mea
ls an
d be
dtim
e sn
ack.
Reas
ons
for
with
draw
als:
1 C
SII,
1 M
DI a
fter
4w
eeks
of t
hera
py(r
easo
n no
t sp
ecifi
ed);
1 af
ter
2 w
eeks
and
1 a
fter
2m
onth
s of
bot
hC
SII a
nd M
DI
Con
flict
of i
nter
est:
Boeh
ringh
er M
annh
eim
Can
ada
loan
ed t
here
flect
ance
met
ers
(for
BG m
easu
rem
ent)
. Fun
ded
by D
iabe
ticC
hild
ren’
s Fo
unda
tion,
Que
bec
and
the
Mon
trea
l Chi
ldre
n’s
Hos
pita
lRe
sear
ch In
stitu
te
Popu
latio
n: in
sulin
-dep
ende
nt d
iabe
tics
with
sev
ere
insu
lin d
efic
ienc
y do
cum
ente
dby
abs
ent
C-p
eptid
e re
spon
se t
o gl
ucag
onst
imul
atio
n. A
ged
14–4
1 y.
Incl
usio
n cr
iteria
: not
rep
orte
d
Part
icip
ants
: N
: 20
End
of s
tudy
: 16
Fem
ale:
70%
Mea
n ag
e (m
in.–
max
.): 2
4.95
(14–
41) y
Dia
bete
s du
ratio
n m
ean
(min
.–m
ax.):
10.3
5 (6
–21)
yBa
selin
e va
lues
: mea
n (±
SD):
HbA
1: 1
3.2
(±1.
1)%
Insu
lin: 4
8 IU
/day
BMI:
22.8
1 kg
/m2
Wei
ght:
61.8
6kg
Prev
ious
reg
imen
: con
vent
iona
l tre
atm
ent
of r
egul
ar a
nd in
term
edia
te-a
ctin
g in
sulin
. 2
inje
ctio
ns/d
ay (n
= 1
6), 3
inje
ctio
ns/d
ay(n
= 3
), 4
inje
ctio
ns/d
ay (n
= 1
)
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
121
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Schm
itz e
t al
.,19
8986
Den
mar
k, s
ingl
e-ce
ntre
,ra
ndom
ised
cros
sove
r
Dur
atio
n: C
SII
6m
onth
s, M
DI
6m
onth
s
Trea
tmen
t 1:
CSI
I+
sol
uble
.N
ordi
sk
Trea
tmen
t 2:
MD
I+
reg
ular
(3–4
×so
lubl
e, 1
×N
PH.
Insu
ject
pen
+)
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
of o
utco
mes
:M
etho
d of
rec
ordi
ng h
ypos
not
reco
rded
. Pts
cho
seth
eir
pref
eren
tial
trea
tmen
t at
end
of s
tudy
;m
etho
d no
t re
cord
ed
Allo
catio
n to
tre
atm
ent
grou
ps: p
ts r
ando
mise
d, b
ut n
o in
dica
tion
ofm
etho
d of
ran
dom
isatio
n. A
lloca
tion
conc
ealm
ent
uncl
ear
Run-
in/w
ash-
out
perio
d: n
ot r
epor
ted
Com
para
bilit
y of
tre
atm
ent
grou
ps: n
ot r
epor
ted
Ana
lysis
: Mea
ns ±
assu
me
SD (n
ot s
peci
fied)
. Diff
eren
ces
judg
ed b
ySt
uden
t’s p
aire
d t-
test
. HbA
1cm
easu
red
at 0
mon
ths
and
mon
thly
and
mea
n of
val
ues
durin
g la
st 3
mon
ths
of e
ach
stud
y pe
riod
calc
ulat
ed. W
eigh
t an
d in
sulin
dos
e ca
lcul
ated
as
mea
n of
val
ues
durin
g 3
mon
ths
prio
r to
stu
dy a
nd d
urin
g la
st 3
mon
ths
of e
ach
stud
y pe
riod
Sam
ple
size/
pow
er c
alcu
latio
n: n
ot r
epor
ted
Gen
eral
isabi
lity:
adu
lts w
ith lo
ng-t
erm
unc
ompl
icat
ed in
sulin
-de
pend
ent
diab
etes
. In
good
con
trol
bef
ore
entr
ance
to
stud
yO
utco
me
mea
sure
s: p
ts s
een
mon
thly
in o
utpa
tient
clin
ic. V
ery
little
info
rmat
ion
on h
ypos
– o
nly
repo
rts
that
‘no
seve
re h
ypos
occ
urre
d’C
onfli
ct o
f int
eres
t: no
ne r
epor
ted
Popu
latio
n: o
ther
wise
hea
lthy
insu
lin-
depe
nden
t di
abet
ics,
age
d 26
–50
y.D
iabe
tes
in ‘g
ood
cont
rol’
befo
re e
ntra
nce
to t
he s
tudy
Incl
usio
n cr
iteria
: dia
bete
s du
ratio
n ≥
20y,
abse
nce
of s
ever
e di
abet
ic c
ompl
icat
ions
,i.e
. urin
ary
albu
min
exc
retio
n ra
te≤
20µg
/min
, nor
mal
blo
od p
ress
ure
and
seru
m c
reat
inin
e an
d no
pro
lifer
ativ
ere
tinop
athy
. Also
, no
med
icat
ion
othe
rth
an in
sulin
tak
en
Part
icip
ants
: N
: 10
End
of s
tudy
: 10
Fem
ale:
60%
Mea
n ag
e (±
SD):
36.5
(±7.
9) y
, min
. 26,
max
. 50
Dia
bete
s du
ratio
n m
ean
(±SD
): 23
.7(±
2.9)
y, m
in. 2
1, m
ax. 3
0Ba
selin
e va
lues
: mea
n (±
SD):
HbA
1c: 7
.6 (±
0.9)
%
Insu
lin: 0
.6 (±
0.11
) IU
/kg/
day
BMI:
23kg
/m2
(cal
cula
ted
by r
evie
wer
)W
eigh
t: 71
.6 (±
17.2
)kg
Prev
ious
reg
imen
: con
vent
iona
l tre
atm
ent
(no
furt
her
deta
ils)
cont
inue
d
Appendix 6
122
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Tsui
et
al.,
2001
79
Can
ada,
sin
gle-
cent
re, R
CT
(par
alle
l)
Dur
atio
n:9
mon
ths
Trea
tmen
t 1:
CSI
I+
lisp
ro. M
inim
ed50
7
Trea
tmen
t 2:
MD
I+
lisp
ro (a
lso 1
×N
PH).
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
ofou
tcom
es: A
dver
seev
ents
suc
h as
DK
Are
cord
ed b
y pa
tient
s.H
bA1c
mea
sure
dm
onth
ly.H
ypog
lyca
emia
:sy
mpt
oms
relie
ved
byin
gest
ion
of g
luco
sean
d/or
BG
≤3
mm
ol/l;
seve
re h
ypog
lyca
emia
:re
quire
s as
sista
nce
orin
com
a
Allo
catio
n to
tre
atm
ent
grou
ps: p
ts r
ando
mise
d af
ter
a 2-
wee
k sc
reen
ing
perio
d by
sea
led
enve
lope
s, p
repa
red
inde
pend
ently
usin
g co
mpu
ter-
gene
rate
d ra
ndom
isatio
nsc
hedu
leRu
n-in
/was
h-ou
t pe
riod:
not
rep
orte
dC
ompa
rabi
lity
of t
reat
men
t gr
oups
: pts
sim
ilar
at b
asel
ine
Diff
eren
ce in
HbA
1c: n
ot s
tatis
tical
ly s
igni
fican
tA
naly
sis: H
bA1c
anal
ysed
by
rand
om-e
ffect
s re
gres
sion
mod
el; m
onth
ly h
ypos
ana
lyse
d by
Poi
sson
reg
ress
ion,
base
line
HbA
1cus
ed a
s a
cova
riate
in m
odel
. Mul
tivar
iate
linea
r m
odel
ling
used
for
subs
cale
of D
QO
L an
d m
ultiv
aria
tean
alys
is of
var
ianc
e gl
obal
ly, e
ach
trea
tmen
t gr
oup
test
ed o
nits
ow
n an
d w
ith a
ge, s
ex, d
iseas
e du
ratio
n, b
asel
ine
HbA
1c,
tota
l no.
of h
ypos
, and
no.
of s
ever
e hy
pos
incl
uded
sepa
rate
ly a
nd c
ombi
ned
as c
ovar
iate
s. 9
5% C
I and
p-v
alue
sca
lcul
ated
. p-v
alue
of 0
.05
used
as
signi
fican
ce, n
ot a
djus
ted
for
mul
tiple
com
paris
ons.
Pts
ana
lyse
d ac
cord
ing
tora
ndom
isatio
n gr
oup
Sam
ple
size/
pow
er c
alcu
latio
n: 2
8 pt
s re
crui
ted
but
only
27
rand
omise
d –
1 w
as u
nsui
tabl
e. S
tatis
tical
pow
er o
f 86%
to
dete
ct a
clin
ical
ly im
port
ant
over
all d
iffer
ence
of ±
0.5
inH
bA1c
betw
een
two
trea
tmen
t gr
oups
usin
g tw
o-ta
iled
alph
aof
0.0
5G
ener
alisa
bilit
y: a
dults
age
d 18
–60
y w
ith in
sulin
-dep
ende
ntdi
abet
esO
utco
me
mea
sure
s: r
ecor
ded
adve
rse
even
ts s
uch
as D
KA
,H
bA1c
, mon
thly
Oth
er: A
t en
d of
9m
onth
s st
udy
12 M
DI c
ross
ed t
o C
SII a
ndfo
llow
ed fo
r 6
mon
ths.
Firs
t 3
mon
ths
excl
uded
from
anal
ysis,
but
stu
dy d
one
post
hoc
and
pote
ntia
l for
bia
s du
e to
perio
d ef
fect
Reas
ons
for
with
draw
als:
1 C
SII d
ropp
ed o
ut a
fter
3m
onth
s(r
easo
n no
t gi
ven)
Con
flict
of i
nter
est:
1 au
thor
rec
eive
d re
sear
ch s
uppo
rt,
hono
raria
and
con
sulta
nt fe
es fr
om E
li Li
lly, o
ne a
utho
r is
ace
rtifi
ed t
rain
er fr
om M
iniM
ed. S
tudy
sup
port
ed b
y re
sear
chgr
ant
from
Eli
Lilly
. Pum
ps a
nd d
ispos
able
sup
plie
s by
Min
iMed
. Hum
alog
and
NPH
by
Eli L
illy
Popu
latio
n: a
dults
age
d 18
–60
y w
ith e
ndoc
rine
diag
nose
d Ty
pe I
diab
etes
Incl
usio
n cr
iteria
: dia
betic
for
>2
y, on
set o
f dia
bete
s on
or b
efor
e 40
y, ab
le to
com
ply
with
trea
tmen
t reg
imen
,cu
rren
tly r
ecei
ving
two
or m
ore
inje
ctio
ns p
er d
ay a
ndin
tere
sted
and
mot
ivat
ed to
use
CSI
IEx
clus
ion:
hist
ory
of >
2 se
vere
hyp
os (c
oma,
sei
zure
,lo
ss o
f con
scio
usne
ss) i
n th
e la
st y
ear,
haem
oglo
bino
path
y, in
sulin
res
istan
ce, e
xtre
me
obes
ity(B
MI>
35),
seve
re la
te c
ompl
icat
ions
of d
iabe
tes,
evid
ence
of s
igni
fican
t car
diov
ascu
lar,
hepa
tic d
iseas
e,ca
ncer
, or
cere
brov
ascu
lar
or s
ever
e pe
riphe
ral v
ascu
lar
dise
ase,
alc
ohol
or
drug
abu
se, a
nd/o
r pa
rtic
ipat
ion
inan
othe
r cl
inic
al tr
ial i
n pa
st 4
wee
ks. P
regn
ant w
omen
or
thos
e lik
ely
to b
ecom
e pr
egna
nt. 1
exc
lude
d as
too
hirs
ute
for
pum
p ad
hesiv
e to
stic
kC
SII p
artic
ipan
ts:
N: 1
3En
d of
stu
dy: 1
2Fe
mal
e: 3
8%M
ean
age
(±SD
): 36
(±12
) yrs
Med
ian
age
(min
-max
): 38
(19-
57)y
Dia
bete
s du
ratio
n m
ean
(±SD
): 17
(±10
), m
in. 8
, max
.37
yBa
selin
e va
lues
: mea
n (±
SD):
HbA
1c: 7
.73
(±0.
6)%
In
sulin
: 0.7
(±0.
2) IU
/day
BMI:
27 (±
4) k
g/m
2
MD
I par
ticip
ants
:N
: 14
End
of s
tudy
: 14
Fem
ale:
29%
Mea
n ag
e (±
SD):
36 (±
10) y
Med
ian
age
(min
-max
): 35
(21–
58) y
Dia
bete
s du
ratio
n m
ean
(±SD
): 15
(±9)
, min
. 4, m
ax.
28 y
rsBa
selin
e va
lues
: mea
n (±
SD):
HbA
1c: 8
.16
(±0.
7)%
, 95
% C
I –0.
94 to
+0.
09, p
>0.
10.
Insu
lin: 0
.7 (±
0.1)
IU/d
ayBM
I: 26
(±3)
kg/
m2
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
123
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Zie
gler
et
al.,
1990
84
Ger
man
y, s
ingl
e-ce
ntre
, RC
T(p
aral
lel)
Dur
atio
n:24
mon
ths
Trea
tmen
t 1:
CSI
I+
sol
uble
.N
ordi
sk In
fuse
r,Be
tatr
on I
and
II,au
tosy
ringe
8M
P,or
Pro
med
os E
1
Trea
tmen
t 2:
MD
I+
reg
ular
(2–4
inje
ctio
ns d
aily
: 2in
ject
ions
(sta
ndar
d) =
com
bina
tions
of
regu
lar
and
NPH
.3
inje
ctio
ns(in
crea
sed
fast
ing
BG) =
2 ×
NPH
+ r
egul
ar, a
nd
2–3
×re
gula
r. 4
inje
ctio
ns(ir
regu
lar
lifes
tyle
)=
1–2
×N
PH a
nd3–
4 ×
regu
lar)
.
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
ofou
tcom
es: s
yste
mat
icas
sess
men
t of
hyp
o-an
d hy
perg
lyca
emic
episo
des
base
d on
pts
’da
ily r
ecor
ds a
nd o
nin
quiri
es a
t ea
chm
onth
ly c
linic
visi
t.M
ild h
ypo
= B
G<
50m
g/dl
with
out
sym
ptom
s, o
r th
eap
pear
ance
of t
ypic
alsy
mpt
oms
that
cou
ldbe
sel
f-m
anag
ed.
Seve
re h
ypo
= e
vent
sre
quiri
ng a
ssist
ance
by
othe
r pe
rson
s, o
rco
ma
whe
n gl
ucos
e or
gluc
agon
inje
ctio
nsw
ere
requ
ired.
Hyp
ergl
ycae
mic
keto
sis =
BG
>20
0m
g/dl
and
posit
ive
urin
e ac
eton
e.Ke
toac
idos
is =
ven
ous
bloo
d pH
<7.
3 at
hosp
ital a
dmiss
ion.
HbA
1de
term
ined
mon
thly
Allo
catio
n to
tre
atm
ent
grou
ps: p
ts r
ando
mise
d, b
ut n
oin
dica
tion
of m
etho
d of
ran
dom
isatio
n. A
lloca
tion
conc
ealm
ent
uncl
ear
Run-
in/w
ash-
out
perio
d: fo
r 1
mon
th p
rior
to r
ando
misa
tion,
all p
ts w
ere
put
on in
tens
ified
con
vent
iona
l inj
ectio
n th
erap
y.A
lso 1
-wee
k te
achi
ng p
rogr
amm
eC
ompa
rabi
lity
of t
reat
men
t gr
oups
: clin
ical
cha
ract
erist
ics
atra
ndom
isatio
n gi
ven
for
both
gro
ups.
Sim
ilar
age,
BM
I,du
ratio
n of
dia
bete
s, in
sulin
dos
e an
d H
bA1.
Ana
lysis
: met
hod
stat
es d
ata
are
mea
n ±
SD, b
ut g
raph
s ar
eSE
. Adv
erse
effe
cts
are
mea
n (r
ange
). Ba
selin
e da
ta a
rem
edia
n (m
in.–
max
.). S
tude
nt’s
t-t
est
(2-s
ided
) for
inde
pend
ent
grou
ps. F
isher
’s e
xact
pro
babi
lity
test
(2-s
ided
)to
ana
lyse
qua
litat
ive
varia
bles
. Rev
iew
er h
as e
xtra
pola
ted
grap
h fo
r al
l HbA
1va
lues
. Whe
re p
ts c
hang
ed t
reat
men
tgr
oup,
the
y w
ere
reta
ined
in t
he o
rigin
ally
ass
igne
d gr
oup
for
stat
istic
al a
naly
sisSa
mpl
e siz
e/po
wer
cal
cula
tion:
not
rep
orte
dG
ener
alisa
bilit
y: p
oorly
con
trol
led
Type
1 p
ts w
ith c
linic
alsy
mpt
oms
of a
t le
ast
1 ch
roni
c di
abet
ic c
ompl
icat
ion
Out
com
e m
easu
res:
HbA
1ex
amin
ed a
t ba
selin
e an
d m
onth
lyth
erea
fter.
No.
of a
dver
se e
vent
s ca
lcul
ated
for
each
pt
per
6m
onth
s. T
otal
no.
of s
ever
e hy
po a
nd D
KA
per
24
mon
ths
extr
apol
ated
to
100
y (n
o ev
ents
/100
y).
Reas
ons
for
with
draw
al: p
artic
ipat
ion
offe
red
to 2
08, 1
08re
fuse
d du
e to
stu
dy p
roto
col o
r di
stan
ce t
o ce
ntre
. 4ex
clud
ed a
t en
d of
rec
ruitm
ent
phas
e as
did
not
mee
tcr
iteria
. With
draw
als:
adv
erse
effe
cts
(not
spe
cifie
d) (1
CSI
I);re
fusa
l to
cont
inue
on
ther
apy
(1 C
SII,
1 M
DI);
bur
den
ofst
udy
(e.g
. reg
ular
BG
sel
f-m
onito
ring)
(6 C
SII,
5 M
DI);
reas
ons
unre
late
d to
stu
dy (e
.g. m
oved
hou
se) (
5 C
SII,
4M
DI)
Oth
er: 1
1 pt
s (4
CSI
I, 7
MD
I) ch
ange
d to
oth
er t
reat
men
t. In
both
gro
ups
chan
gers
equ
ally
dist
ribut
ed t
hrou
ghou
t st
udy
perio
d. 1
pt
with
aut
onom
ic n
euro
path
y di
ed fr
om s
udde
nde
ath
with
out
evid
ence
of h
ypo
of D
KA
– a
ssum
ed t
hey
are
incl
uded
in t
he g
roup
of w
ithdr
awal
s ‘u
nrel
ated
to
stud
y’C
onfli
ct o
f int
eres
t: su
ppor
ted
by g
rant
s fr
om D
euts
che
Fors
chun
gsge
mei
nsch
aft,
Bonn
FRG
(SFB
113
), th
eBu
ndes
min
ister
für
Ges
undh
eit
and
the
Min
ister
für
Wiss
ensc
haft
und
Fors
chun
g, N
RW
Popu
latio
n: C
-pep
tide
–ve,
poo
rly c
ontr
olle
d Ty
pe 1
diab
etes
pts
with
clin
ical
sym
ptom
s of
at
leas
t 1
chro
nic
diab
etic
com
plic
atio
n, w
ho w
ere
unse
lect
ed a
sto
inte
llige
nce
and
educ
atio
n. A
ged
18–6
0 y.
Incl
usio
n cr
iteria
: ≥18
y an
d ≤
60 y
; Typ
e 1
diab
etes
;m
anife
stat
ion
of d
iabe
tes
<35
y of
age
; dia
bete
sdu
ratio
n >
1y,
stim
ulat
ed s
erum
C-p
eptid
e<
0.3
ng/m
l; H
bA1
≥8.
0% u
nder
con
vent
iona
l ins
ulin
ther
apy;
at
leas
t 1
chro
nic
diab
etic
com
plic
atio
n(r
etin
opat
hy, n
ephr
opat
hy o
r ne
urop
athy
)Ex
clus
ion:
mal
igna
ncie
s; p
revi
ous
card
iac
infa
rctio
n;en
docr
ine
diso
rder
s; p
rimar
y dy
slipo
prot
eina
emia
;am
auro
sis; h
ypog
lyca
emic
una
war
enes
s; s
erum
crea
tinin
e >
3m
g/dl
; psy
chos
is; c
hron
ic a
lcoh
ol a
buse
;in
form
ed c
onse
nt r
efus
edC
SII p
artic
ipan
ts:
N: 4
9En
d of
stu
dy: 3
6Fe
mal
e: 3
8.8%
Med
ian
age
(min
.–m
ax.):
32
(18–
54) y
Dia
bete
s du
ratio
n m
edia
n (m
in.–
max
.): 1
8.0
(3–4
4) y
Base
line
valu
es: m
ean
(±SD
):H
bA1:
9.8
%
HbA
11
mon
th p
rior
to r
ando
misa
tion
10.6
(±2.
2)In
sulin
: 49
(±17
) IU
/day
BMI:
23 k
g/m
2
MD
I par
ticip
ants
:N
: 47
End
of s
tudy
: 37
Fem
ale:
51.
1%M
edia
n ag
e (m
in.–
max
.): 3
2 (1
8–55
) yD
iabe
tes
dura
tion
med
ian
(min
.–m
ax.):
14.
8 (3
–42)
yBa
selin
e va
lues
: mea
n (±
SD):
HbA
1: 9
.5%
, p=
ns
HbA
11
mon
th p
rior
to r
ando
misa
tion
10.2
(±1.
6),
p=
nsIn
sulin
: 48
(±14
) IU
/day
, p=
nsBM
I: 22
.1 k
g/m
2
Prev
ious
reg
imen
: con
vent
iona
l ins
ulin
the
rapy
, def
ined
as d
aily
inje
ctio
ns o
f lon
g-ac
ting
or fi
xed
com
bina
tions
of in
term
edia
te-
and
shor
t-ac
ting
insu
lin a
t gi
ven
dose
sw
ithou
t re
gula
r BG
sel
f-m
onito
ring
and
self-
adju
stm
ent
of t
hera
py
AN
OVA
, ana
lysis
of v
aria
nce.
Health Technology Assessment 2004; Vol. 8: No. 43
125
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Appendix 7
Summary of methodology: pregnancy
Appendix 7
126
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Burk
art
et a
l.,19
8895
Ger
man
y, s
ingl
e-ce
ntre
, RC
T(p
aral
lel)
Dur
atio
n:9
mon
ths
(Also
com
pare
dw
ith a
conv
entio
nal
trea
tmen
t gr
oup
– da
ta n
otex
trac
ted)
Trea
tmen
t 1:
CSI
I+
sol
uble
. ‘O
pen
loop
infu
sion
syst
em’ –
no
furt
her
deta
ils
Trea
tmen
t 2:
MD
I+
reg
ular
.D
escr
ibed
as
‘inte
nsifi
ed in
sulin
ther
apy’
by
‘con
vent
iona
lsy
ringe
tec
hniq
ue’.
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
ofou
tcom
es:
Dat
a pr
esen
ted
acco
rdin
g to
sev
erity
of m
ater
nal d
iabe
tes
(Whi
te’s
cla
ssifi
catio
n).
Mea
n bi
rthw
eigh
t an
dge
stat
iona
l age
for
each
gro
up e
stim
ated
from
the
mea
nsre
port
ed fo
r ea
ch o
fW
hite
’s c
lass
ifica
tion.
Feta
l mor
bidi
ty:
heal
thy
new
born
s –
i.v. g
luco
se <
48 h
,1
min
or s
ympt
om;
min
or s
ympt
oms
– i.v
.gl
ucos
e >
48 h
,<
5da
ys, o
ne o
r m
ore
of p
olyc
ytha
emia
,hy
perb
iliru
bina
emia
,hy
poca
lcae
mia
, lig
htre
spira
tory
dist
ress
synd
rom
e, b
irthw
eigh
t>
90th
per
cent
ile;
maj
or s
ympt
oms
– 2
or m
ore
sym
ptom
s of
birt
hwei
ght
>95
thpe
rcen
tile,
card
iom
yopa
thy,
hepa
tosp
leno
meg
aly,
seve
re r
espi
rato
rydi
stre
ss s
yndr
ome,
spas
ms,
i.v.
glu
cose
usua
lly >
5 da
ys.
Prem
atur
e -
deliv
ery
befo
re e
nd o
fge
stat
iona
l wee
k 37
Allo
catio
n to
tre
atm
ent
grou
ps: p
atie
nts
‘rand
omly
sel
ecte
dto
eac
h of
the
pro
cedu
res’
. No
furt
her
deta
ils. N
o de
tails
of
allo
catio
n co
ncea
lmen
tRu
n-in
-was
h ou
t pe
riod:
not
app
licab
leC
ompa
rabi
lity
of t
reat
men
t gr
oups
: age
and
dur
atio
n of
diab
etes
sim
ilar.
CSI
I had
mor
e cl
ass
D (p
<0.
05) a
nd le
sscl
ass
B (p
<0.
005)
tha
n M
DI (
Whi
te’s
cla
ssifi
catio
n).
Sim
ilarit
y of
bas
elin
e H
bA1
and
othe
r va
riabl
es n
ot r
epor
ted
Ana
lysis
: Sta
tes
‘rout
ine
stat
istic
al m
etho
ds’ u
sed.
No
furt
her
deta
ils. U
nits
for
birt
hwei
ght:
not
give
nSa
mpl
e siz
e/po
wer
cal
cula
tion:
not
rep
orte
dG
ener
alisa
bilit
y: p
regn
ant
wom
en w
ith T
ype
1 di
abet
esO
utco
me
mea
sure
s: p
regn
ancy
com
plic
atio
ns, f
etal
out
com
e.Bl
ood
gluc
ose,
HbA
1an
d in
sulin
dos
e no
t re
port
edC
onfli
ct o
f int
eres
t: no
ne s
tate
d
Popu
latio
n: p
regn
ant
wom
en w
ith T
ype
1 di
abet
esw
ho a
tten
ded
the
hosp
ital n
o la
ter
than
the
firs
ttr
imes
ter
Excl
usio
n cr
iteria
: Typ
e 2
and
gest
atio
nal d
iabe
tes,
patie
nts
who
att
ende
d th
e ho
spita
l lat
er t
han
in t
hefir
st t
rimes
ter
CSI
I par
ticip
ants
:N
: 48
End
of s
tudy
: 48
Fem
ale:
100
%M
ean
age
(±SD
): 28
.4 (±
5.3)
yD
iabe
tes
dura
tion
mea
n (±
SD):
11.7
(±7.
4)y
MD
I par
ticip
ants
:N
: 41
End
of s
tudy
: 41
Fem
ale:
100
%M
ean
age
(±SD
): 28
.7 (±
4.7)
yD
iabe
tes
dura
tion
mea
n (±
SD):
9.9
(±8.
9) y
Base
line
valu
es: n
ot g
iven
. Sta
tes
that
pat
ient
s se
lect
edin
the
CSI
I or
MD
I gro
up h
ad n
orm
al m
ean
gluc
ose
and
HbA
1<
7.5%
at
leas
t fr
om t
he e
nd o
f the
firs
ttr
imes
ter
on.
Prev
ious
reg
imen
: not
sta
ted
for
eith
er g
roup
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
127
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Car
ta e
t al
.,19
8694
Italy,
sin
gle-
cent
re, R
CT
(par
alle
l)
Dur
atio
n:9
mon
ths
Trea
tmen
t 1:
CSI
I+
sol
uble
.M
icro
jet
MC
20
Trea
tmen
t 2:
MD
I+
reg
ular
(Act
rapi
d ×
4on
ly).
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
ofou
tcom
es: H
bA1
mea
sure
d ea
chtr
imes
ter
Allo
catio
n to
tre
atm
ent
grou
ps: n
o in
dica
tion
of m
etho
d of
rand
omisa
tion
Run-
in/w
ash-
out
perio
d: n
ot a
pplic
able
Com
para
bilit
y of
tre
atm
ent
grou
ps: s
imila
r ag
e. B
asel
ine
HbA
1no
t sig
nific
antly
diff
eren
t fo
r C
SII v
s M
DI a
rms
in b
oth
Type
1 a
nd T
ype
2. B
asel
ine
insu
lin r
equi
rem
ents
not
signi
fican
tly d
iffer
ent
in T
ype
1, b
ut n
ot p
rese
nted
for
Type
2A
naly
sis: s
igni
fican
ce o
f diff
eren
ce b
etw
een
resu
lts a
sses
sed
by m
eans
of 2
-tai
led
Stud
ents
t-t
est
for
unpa
ired
data
.Ba
selin
e H
bA1
and
HbA
1an
d so
me
insu
lin v
alue
s ex
trac
ted
from
figu
re b
y re
view
er. I
ndiv
idua
l pat
ient
bas
elin
e da
tapr
esen
ted,
mea
ns c
alcu
late
d by
rev
iew
erSa
mpl
e siz
e/po
wer
cal
cula
tion:
not
rep
orte
dG
ener
alisa
bilit
y: P
regn
ant
wom
en w
ith T
ype
1 an
d 2
diab
etes
(ana
lyse
d se
para
tely
) who
had
dia
bete
s be
fore
pre
gnan
cyO
utco
me
mea
sure
s: in
sulin
req
uire
men
t, pe
rinat
al o
utco
me
and
glyc
aem
ic c
ontr
ol. W
eekl
y vi
sits
to o
utpa
tient
s w
here
insu
lin d
ose
adju
sted
Con
flict
of i
nter
est:
supp
orte
d by
Gra
nt o
f Reg
ione
Pie
mon
te
Popu
latio
n: p
regn
ant
diab
etic
wom
en –
cla
ssifi
ed a
sTy
pe 1
(15)
and
Typ
e 2
(14)
bef
ore
preg
nanc
yC
riter
ia: n
ot r
epor
ted
Type
I C
SII p
artic
ipan
ts:
N: 8
End
of s
tudy
: 8Fe
mal
e: 1
00%
Mea
n ag
e (m
in.–
max
.): 2
4.6
(17-
37) y
Base
line
valu
es: m
ean
(±SD
):H
bA1:
8.7
5%
Insu
lin: 3
1.5
IU/d
ay
Type
I M
DI p
artic
ipan
ts:
N: 7
End
of s
tudy
: 7Fe
mal
e: 1
00%
Mea
n ag
e (m
in.–
max
.): 2
5.9
(19–
32) y
Base
line
valu
es: m
ean
(±SD
):H
bA1:
9.1
%, p
= n
sIn
sulin
: 27.
8 IU
/day
, p=
ns
Type
1 p
revi
ous
regi
men
: 13
prev
ious
ly o
nco
nven
tiona
l, 2
CSI
I sta
rted
3–4
mon
ths
befo
reco
ncep
tion
Type
2 C
SII p
artic
ipan
ts:
N: 6
End
of s
tudy
: 6Fe
mal
e: 1
00%
Mea
n ag
e (m
in.–
max
.): 3
1.3
(25-
39) y
Base
line
valu
es: m
ean
(±SD
):H
bA1:
8.1
%
Type
2 M
DI p
artic
ipan
ts:
N: 8
End
of s
tudy
: 8Fe
mal
e: 1
00%
Mea
n ag
e (m
in.–
max
.): 2
9.4
(20–
46) y
Base
line
valu
es: m
ean
(±SD
):H
bA1:
8.2
%, p
= n
s
Type
2 p
revi
ous
regi
men
: 4 t
akin
g or
al h
ypog
lyca
emic
sbe
fore
pre
gnan
cy, 1
0 tr
eate
d by
die
tary
mea
sure
s on
ly
cont
inue
d
Appendix 7
128
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Cou
stan
et
al.,
1986
93
USA
, sin
gle-
cent
re, R
CT
(par
alle
l)
Dur
atio
n:9
mon
ths
Trea
tmen
t 1:
CSI
I+
sol
uble
.A
utos
yrin
ge A
S-2C
, 6C
,6C
(U10
0); L
illy
CPI
-910
0
Trea
tmen
t 2:
MD
I+
reg
ular
(2–4
inte
rmed
iate
or
rapi
d-ac
ting
insu
lin)
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
ofou
tcom
es: s
ubje
ctiv
ehy
pos
reco
rded
indi
ary:
mild
(res
pond
edto
sel
f-ad
min
ister
edsn
ack)
, mod
erat
e(r
equi
ring
anot
her
pers
on’s
ass
istan
ce),
seve
re (r
equi
red
gluc
agon
or
intr
aven
ous
gluc
ose,
usua
lly in
em
erge
ncy
room
set
ting)
. HbA
1m
easu
red
at w
eekl
ycl
inic
visi
ts. P
tssw
itche
d to
cont
inuo
usin
trav
enou
s gl
ucos
ean
d in
sulin
infu
sion
whe
n ad
mitt
ed t
ola
bour
, ele
ctiv
e la
bour
or C
aesa
rean
sec
tion
Allo
catio
n to
tre
atm
ent
grou
ps: p
ts r
ando
mise
d us
ing
seal
eden
velo
pes.
The
sub
ject
stu
died
dur
ing
2 pr
egna
ncie
s w
asra
ndom
ised
inde
pend
ently
eac
h tim
e an
d tr
eate
d w
ith C
SII
durin
g on
e pr
egna
ncy
and
MD
I dur
ing
othe
rRu
n-in
/was
h-ou
t pe
riod:
48
h on
exi
stin
g re
gim
en a
sin
patie
nt fo
r ba
selin
e da
ta. A
fter
rand
omisa
tion
rem
aine
d in
inpa
tient
for
up t
o 7
days
to
optim
ise c
ontr
ol o
n as
signe
dtr
eatm
ent
Com
para
bilit
y of
tre
atm
ent
grou
ps: t
able
of b
asel
ine
char
acte
ristic
s sh
ows
grou
ps t
o be
com
para
ble,
rep
orts
no
signi
fican
t di
ffere
nces
.A
naly
sis: m
eans
and
SD
pre
sent
ed. S
tude
nt’s
t-t
est
(pai
red
and
unpa
ired)
. Wilc
oxon
non
-pai
red
rank
-sum
tes
t. H
bA1
mon
th 2
dat
a ex
trac
ted
from
figu
reSa
mpl
e siz
e/po
wer
cal
cula
tion:
sam
ple
size
22 p
regn
anci
esal
low
ing
dete
ctio
n of
a d
iffer
ence
in a
vera
ge c
ircul
atin
ggl
ucos
e le
vels
betw
een
the
two
grou
ps o
f 17.
9m
g/dl
at
0.05
leve
l of s
igni
fican
ce a
nd 8
0% p
ower
, ass
umin
g an
SD
of
15m
g/dl
. Gro
ups
too
smal
l for
sta
tistic
ally
sig
nific
ant
diffe
renc
e in
mod
erat
e an
d se
vere
hyp
os t
o be
att
aine
d –
with
a p
ower
of 8
0% if
50%
in o
ne g
roup
and
90%
in t
heot
her
grou
p re
port
ed s
ympt
oms
29 p
atie
nts
wou
ld b
ere
quire
d in
eac
h gr
oup
Gen
eral
isabi
lity:
pre
gnan
t in
sulin
dep
ende
nt w
omen
Out
com
e m
easu
res:
HbA
1re
port
ed. H
ypos
rec
orde
d by
patie
nts.
Con
flict
of i
nter
est:
stud
y su
ppor
ted
by g
rant
6-2
60 fr
om t
heM
arch
of D
imes
Birt
h D
efec
ts F
ound
atio
n, g
rant
s RR
125
and
AM
2049
5 fr
om t
he N
IH a
nd g
rant
s fr
om t
he Ju
veni
leD
iabe
tes
Foun
datio
n, A
mer
ican
Dia
bete
s A
ssoc
iatio
n an
d th
eD
iabe
tes
Ass
ocia
tion
of G
reat
er F
all R
iver
, Mas
sach
uset
ts.
Insu
lin p
rovi
ded
by S
quib
b-N
ovo
Popu
latio
n: in
sulin
-dep
ende
nt d
iabe
tes.
Wom
en –
preg
nant
(15)
or
plan
ning
pre
gnan
cy (7
). 21
wom
enw
ith 2
2 pr
egna
ncie
sC
riter
ia: n
ot s
peci
fied
CSI
I par
ticip
ants
: N
: 11
End
of s
tudy
: 11
Fem
ale:
100
%M
ean
age
(±SD
): 29
(±3)
yrs
, min
. 26,
max
. 32
Base
line
valu
es: m
ean
(±SD
):H
bA1:
8.6
(±1.
7) %
In
sulin
: 0.9
(±0.
42) I
U/d
ay%
obe
se (>
20%
ove
r id
eal b
ody
wei
ght)
= 2
MD
I par
ticip
ants
:N
: 11
End
of s
tudy
: 11
Fem
ale:
100
%M
ean
age
(±SD
): 28
(±4)
y, m
in. 2
4, m
ax. 3
2Ba
selin
e va
lues
: mea
n (±
SD):
HbA
1: 9
.1 (±
1.5)
% ,
p=
ns
Insu
lin: 1
.05
(±0.
38) I
U/d
ay, p
= 0
.381
% o
bese
(>20
% o
ver
idea
l bod
y w
eigh
t) =
2
Prev
ious
reg
imen
: at
leas
t 2
daily
inje
ctio
ns o
f am
ixtu
re o
f rap
id-
and
inte
rmed
iate
-act
ing
insu
lins
(100
%)
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
129
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Nos
ari e
t al
.,19
9392
Italy,
sin
gle-
cent
re, R
CT
(par
alle
l)
Dur
atio
n:9
mon
ths
Trea
tmen
t 1:
CSI
I+
sol
uble
.M
icro
jet
20 o
rD
ahed
i BV
Trea
tmen
t 2:
MD
I+
reg
ular
(3 ×
solu
ble,
1 ×
inte
rmed
iate
)
Reas
on fo
r C
SII:
refe
rred
for
inte
nsiv
etr
eatm
ent
beca
use
preg
nant
(28)
or
plan
ning
to
beco
me
preg
nant
(4)
Ass
essm
ent
ofou
tcom
es: h
ypos
clas
sifie
d as
sev
ere
whe
n ch
arac
teris
ed b
yco
ma
or s
eizu
re o
rre
quiri
ngho
spita
lisat
ion
orin
trav
enou
s gl
ucos
e or
gluc
agon
Allo
catio
n to
tre
atm
ent
grou
ps: r
ando
mly
allo
cate
d us
ing
seal
ed e
nvel
opes
. No
furt
her
info
rmat
ion
repo
rted
Run-
in/w
ash-
out
perio
d: n
ot a
pplic
able
Com
para
bilit
y of
tre
atm
ent
grou
ps: s
imila
r ag
e an
d BM
I, bu
tba
selin
e H
bA1c
, mea
n in
sulin
dos
e, d
urat
ion
of p
regn
ancy
of
rand
omisa
tion
not
repo
rted
Ana
lysis
: stu
dent
’s t
-tes
t or
χ2
anal
ysis
Sam
ple
size/
pow
er c
alcu
latio
n: n
ot r
epor
ted
Gen
eral
isabi
lity:
pre
gnan
t w
omen
with
Typ
e 1
diab
etes
Out
com
e m
easu
res:
HbA
1cre
port
ed. B
irthw
eigh
t as
sess
ed in
rela
tion
to g
esta
tiona
l age
, acc
ordi
ng t
o th
e na
tiona
lno
mog
ram
of i
ntra
uter
ine
feta
l gro
wth
Oth
er: 2
8 w
omen
wer
e pr
egna
nt w
hen
refe
rred
, but
dura
tion
of p
regn
ancy
whe
n ra
ndom
ised
not
repo
rted
Con
flict
of i
nter
est:
none
sta
ted
Popu
latio
n: c
lass
ified
with
Typ
e 1
diab
etes
bef
ore
preg
nanc
y. 3
1 w
omen
with
32
preg
nanc
ies
(16
in e
ach
arm
) inc
lude
d.C
riter
ia: n
ot r
epor
ted
CSI
I par
ticip
ants
: N
: 16
End
of s
tudy
: 16
Mea
n ag
e (±
SD):
25.5
(±7.
2) y
Dia
bete
s du
ratio
n: m
in. 3
.0, m
ax. 1
6.0
yBa
selin
e va
lues
: mea
n (±
SD):
BMI:
21.8
(±1.
6)kg
/m2
MD
I par
ticip
ants
:N
: 16
End
of s
tudy
: 16
Mea
n ag
e (±
SD):
27.3
(±12
) yD
iabe
tes
dura
tion:
min
. 3.0
, max
. 16.
0 y
Base
line
valu
es: m
ean
(±SD
):BM
I: 21
.6 (±
2.4)
kg/m
2
Prev
ious
reg
imen
: one
sta
rted
CSI
I 2–4
mon
ths
prio
rto
con
cept
ion,
15
had
at le
ast
2 da
ily in
ject
ions
Health Technology Assessment 2004; Vol. 8: No. 43
131
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Appendix 8
Summary of methodology: adolescents
Appendix 8
132
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Schi
ffrin
et
al.,
1984
97
Can
ada,
sin
gle-
cent
re,
rand
omise
dcr
osso
ver
[3 a
rms
– M
DI,
CSI
I, or
CSI
I at
nigh
t w
ithpr
epra
ndia
lsh
ort-
actin
gin
sulin
dur
ing
the
day
(CSI
I +M
DI)]
Dur
atio
n: C
SII 4
mon
ths,
MD
I4
mon
ths,
CSI
I+M
DI
4m
onth
s
Trea
tmen
t 1:
CSI
I+
sol
uble
. Mill
Hill
2703
Trea
tmen
t 2:
MD
I+
reg
ular
(3–4
×so
lubl
e, 1
×N
PH).
Trea
tmen
t 3:
CSI
I+
MD
I. C
SII
over
nigh
t w
ithpr
epra
ndia
lin
ject
ions
of s
hort
-ac
ting
insu
lindu
ring
the
day.
Pum
p us
ed fo
rpr
ebre
akfa
st b
olus
Oth
erin
terv
entio
ns: p
tsfo
llow
ed t
heC
anad
ian
Dia
bete
sA
ssoc
iatio
n Fo
odch
oice
s di
etsy
stem
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
ofou
tcom
es: o
nce-
mon
thly
fast
ing
bloo
dsa
mpl
es o
btai
ned
for
HbA
1Se
vere
hyp
os –
requ
ired
gluc
agon
Allo
catio
n to
tre
atm
ent
grou
ps: p
ts w
ere
said
to
bera
ndom
ised,
but
no
indi
catio
n of
ran
dom
isatio
n m
etho
d. N
oin
dica
tion
of a
lloca
tion
conc
ealm
ent
Run-
in/w
ash-
out
perio
d: n
one
Com
para
bilit
y of
tre
atm
ent
grou
ps: n
ot r
epor
ted
Ana
lysis
: tre
atm
ent
com
paris
ons
wer
e do
ne b
y 2-
way
AN
OVA
follo
wed
by
the
New
man
–Keu
ls te
st a
nd S
tude
nt’s
t-te
st fo
r pa
ired
data
. IT
T a
naly
sis n
ot c
arrie
d ou
t. H
bA1
data
pres
ente
d in
figu
re o
nly.
HbA
1da
ta e
xtra
cted
from
figu
re b
yre
view
erSa
mpl
e siz
e/po
wer
cal
cula
tion:
not
rep
orte
dG
ener
alisa
bilit
y: a
dole
scen
ts w
ith T
ype
1 di
abet
esO
utco
me
mea
sure
s: H
bA1
repo
rted
Reas
ons
for
with
draw
al: 1
CSI
I afte
r 3
mon
ths
and
1 M
DI
afte
r 1
wee
k (r
easo
n no
t sp
ecifi
ed);
refu
sed
to c
ross
ove
rfr
om C
SII (
n=
2)
Con
flict
of i
nter
est:
gran
t fr
om t
he Ju
veni
le D
iabe
tes
Foun
datio
n an
d th
e C
anad
ian
Dia
bete
s A
ssoc
iatio
n
Popu
latio
n: T
ype
1 pt
s ag
ed 1
3–20
y. P
revi
ously
had
faile
d to
res
pond
to
twic
e-da
ily in
ject
ions
with
hom
eBG
mon
itorin
g re
gim
e. P
ts h
ad n
ot b
een
on in
tens
ive
ther
apy
in t
he p
ast
and
wer
e no
t sig
nific
antly
obe
seC
riter
ia: n
ot r
epor
ted
Part
icip
ants
: N
: 24
End
of s
tudy
: 20
Age
(min
.–m
ax.):
13–
20 y
Dia
bete
s du
ratio
n, m
ean:
9 y
Base
line
valu
es: m
ean
(±SD
):H
bA1:
13
(±1)
%In
sulin
: 64
(±14
) IU
/day
Prev
ious
reg
imen
: tw
ice-
daily
inje
ctio
ns
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
133
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Tam
borla
ne e
tal
., 19
8996
USA
, sin
gle-
cent
re,
rand
omise
dcr
osso
ver
Dur
atio
n: C
SII
6m
onth
s, M
DI
6m
onth
s
Trea
tmen
t 1:
CSI
I+
sol
uble
. Pum
pty
pe n
ot r
epor
ted
Trea
tmen
t 2:
MD
I+
reg
ular
. Ins
ulin
type
and
freq
uenc
y no
tre
port
ed
Reas
on fo
r C
SII:
not
repo
rted
Not
rep
orte
dA
lloca
tion
to t
reat
men
t gr
oups
: ran
dom
ly a
ssig
ned
to C
SII o
rM
DI a
fter
base
line
asse
ssm
ents
obt
aine
d an
d pt
s ad
mitt
ed t
oYa
le C
hild
ren’
s C
linic
al R
esea
rch
Cen
tre.
No
furt
her
deta
ils.
Allo
catio
n co
ncea
lmen
t no
t re
port
edRu
n-in
/was
h-ou
t pe
riod:
pts
rem
aine
d in
hos
pita
l 5–7
day
s to
optim
ise c
ontr
ol o
n as
signe
d tr
eatm
ent
regi
men
. Pts
read
mitt
ed t
o re
sear
ch c
entr
e to
cro
ssov
er t
reat
men
ts. N
ow
ash-
out
perio
d or
ana
lysis
of t
reat
men
t ef
fect
sC
ompa
rabi
lity
of t
reat
men
t gr
oups
: not
rep
orte
dA
naly
sis: d
ata
pres
ente
d as
mea
n (S
E). S
E co
nver
ted
to S
Dby
rev
iew
er. A
naly
sis o
f var
ianc
e us
edSa
mpl
e siz
e/po
wer
cal
cula
tion:
not
rep
orte
dG
ener
alisa
bilit
y: n
on-o
bese
ado
lesc
ents
who
wer
e pr
evio
usly
taki
ng ≤
2 da
ily in
sulin
inje
ctio
ns a
nd n
ot p
erfo
rmin
g fr
eque
ntbl
ood
gluc
ose
mon
itorin
g, t
here
fore
like
ly t
o be
poo
rlyco
ntro
lled
Out
com
e m
easu
res:
HbA
1an
d da
ily in
sulin
dos
e re
port
edD
ropo
uts/
with
draw
als:
not
rep
orte
dC
onfli
ct o
f int
eres
t: su
ppor
ted
by g
rant
s fr
om t
he N
atio
nal
Inst
itute
s of
Hea
lth a
nd t
he Ju
veni
le D
iabe
tes
Foun
datio
nIn
tern
atio
nal
Popu
latio
n: n
on-o
bese
ado
lesc
ents
with
Typ
e 1
diab
etes
,age
d 11
–20
y, h
ad d
iabe
tes
for
2–12
y.
Hei
ghts
wer
e w
ithin
ran
ge o
f nor
mal
with
one
exce
ptio
n. O
ne s
ubje
ct h
ad a
ttai
ned
full
pube
rtal
deve
lopm
ent
Incl
usio
n cr
iteria
: pts
wer
e dr
awn
from
the
Yal
eC
hild
ren’
s D
iabe
tes
Clin
ic a
nd w
ere
incl
uded
if t
hey
wer
e de
emed
to
be a
ble
to c
arry
out
the
tas
ks a
ndre
spon
sibili
ties
dem
ande
d by
the
stu
dy p
roto
col
Excl
usio
n cr
iteria
: pts
tak
ing
mor
e th
an 2
dai
ly in
sulin
inje
ctio
ns o
r pe
rfor
min
g fr
eque
nt s
elf-
bloo
d gl
ucos
em
easu
rem
ents
prio
r to
enr
olm
ent
Part
icip
ants
: N
: 10
Fem
ale:
50%
Age
(min
.–m
ax.):
11–
20 y
Dia
bete
s du
ratio
n (m
in.–
max
.): 2
–12
yBa
selin
e va
lues
: mea
n (±
SD):
HbA
1: 1
2.2
(±4.
1)%
In
sulin
: 1.1
(±0.
32) U
/day
Prev
ious
reg
imen
: con
vent
iona
l ins
ulin
tre
atm
ent
of n
otm
ore
than
2 in
ject
ions
/day
Health Technology Assessment 2004; Vol. 8: No. 43
135
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Appendix 9
Summary of methodology: analogue versus soluble insulin
Appendix 9
136
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Bode
et
al.,
2002
113
USA
, mul
ti-ce
ntre
, RC
T(p
aral
lel)
Dur
atio
n:4
mon
ths
Trea
tmen
t 1:
CSI
I+
lisp
ro
Trea
tmen
t 2:
CSI
I+
buf
fere
d re
gula
r
Trea
tmen
t 3:
CSI
I+
asp
art
Min
imed
/D
isetr
onic
506
or
507
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
ofou
tcom
es:
HbA
1cta
ken
at w
eek
16. T
otal
dai
ly in
sulin
dose
for
wee
k be
fore
base
line
and
last
wee
kof
tre
atm
ent.
Hyp
osy
mpt
oms
and
adve
rse
even
ts r
ecor
ded
indi
ary.
Min
or h
ypo
=sy
mpt
oms
(pal
pita
tions
,tir
edne
ss, s
wea
ting,
stro
ng h
unge
r,di
zzin
ess,
tre
mor
, etc
.)co
nfirm
ed b
y BG
<50
mg/
dl a
nd a
ble
tode
al w
ith it
on
own.
Maj
or h
ypo
= B
G<
50 m
g/dl
and
sev
ere
cent
ral n
ervo
ussy
stem
dys
func
tion
prev
entin
g th
emtr
eatin
g th
emse
lves
or
adm
inist
ratio
n of
pare
nter
al g
luco
se o
rgl
ucag
on
Allo
catio
n to
tre
atm
ent
grou
ps: r
ando
mly
ass
igne
d af
ter
run-
in (2
:2:1
) to
low
est
avai
labl
e ra
ndom
isatio
n nu
mbe
r to
aspa
rt, l
ispro
or
regu
lar.
Rand
omisa
tion
code
pro
vide
d by
Nov
o N
ordi
sk t
o en
sure
tha
t in
vest
igat
or a
nd s
ubje
ct w
ere
blin
ded
at p
oint
of r
ando
misa
tion
Run-
in/w
ash-
out
perio
d: 4
-wee
k ru
n-in
with
buf
fere
d re
gula
r.D
ose
adju
stm
ent
perio
d du
ring
4w
eeks
afte
r ra
ndom
isatio
n.C
ompa
rabi
lity
of t
reat
men
t gr
oups
: pat
ient
s ap
pear
sim
ilar
inag
e, s
ex, B
MI,
HbA
1c, i
nsul
in d
ose
but
stat
istic
s no
tpr
esen
ted
Ana
lysis
: bet
wee
n-tr
eatm
ent
com
paris
ons,
exc
ept
insu
linus
ed, m
ade
with
AN
CO
VA m
odel
with
tre
atm
ent
and
cent
reas
fixe
d ef
fect
s an
d co
rres
pond
ing
base
line
mea
sure
men
t as
cova
riate
. Las
t ob
serv
atio
n ca
rrie
d fo
rwar
d us
ed in
ana
lysis
of H
bA1c
, wei
ght
and
lipid
pro
file.
Sam
ple
size/
pow
er c
alcu
latio
n: n
ot r
epor
ted.
Gen
eral
isabi
lity:
adu
lts w
ith T
ype
1 di
abet
es w
ith a
t le
ast
3m
onth
s C
SII e
xper
ienc
eO
utco
me
mea
sure
s: fi
nal H
bA1c
mea
sure
s an
d SD
/SEM
/CI
not
pres
ente
d, c
hang
es fr
om b
asel
ine
only.
Rat
e of
hyp
os =
mea
n nu
mbe
r (S
D) o
f hyp
os r
epor
ted
by s
ubje
ct p
er 3
0da
ysfo
r al
l sub
ject
s in
the
tre
atm
ent
grou
p. ‘A
ll re
port
ed h
ypos
’ =re
gard
less
of B
G v
alue
. Dat
a ex
trac
ted
for
mai
nten
ance
perio
d on
ly (e
ntire
tria
l and
dos
e-ad
just
men
t pe
riod
valu
esal
so r
epor
ted)
Reas
ons
for
with
draw
al: w
ithdr
awal
of c
onse
nt, l
ack
ofco
mpl
ianc
e, lo
ss t
o fo
llow
-up,
inef
fect
ive
ther
apy
(num
bers
for
each
rea
son
not
spec
ified
): lis
pro
(1),
buffe
red
regu
lar
(9),
aspa
rt (3
); ad
vers
e ev
ent
– he
rpes
zos
ter
(1 a
spar
t).
Con
flict
of i
nter
est:
spon
sore
d by
Nov
o N
ordi
skPh
arm
aceu
tical
s, P
rince
ton,
NJ
Popu
latio
n: a
dults
age
d 18
–71
y w
ith T
ype
1 di
abet
es>
12m
onth
s an
d tr
eate
d w
ith C
SII c
ontin
uous
ly fo
rpr
evio
us 3
mon
ths
Excl
usio
n cr
iteria
: im
paire
d he
patic
, ren
al, o
r ca
rdia
cfu
nctio
n or
rec
urre
nt h
ypog
lyca
emia
. Wom
enpr
egna
nt, b
reas
t-fe
edin
g, o
r no
t us
ing
cont
race
ptio
n
Lisp
ro p
artic
ipan
ts:
N: 2
8En
d of
stu
dy: 2
7Fe
mal
e: 6
8%M
ean
age
(±SD
): 39
.9 (±
11.1
) yBa
selin
e va
lues
: mea
n (±
SD):
HbA
1c: 7
.3 (±
0.7)
%
Insu
lin: 0
.5 (±
0.19
) U/k
gBM
I: 26
.3 (±
3.2)
kg/
m2
Buffe
red
regu
lar
part
icip
ants
:N
: 59
End
of s
tudy
: 50
Fem
ale:
68%
Mea
n ag
e (±
SD):
43.1
(±9.
4) y
Base
line
valu
es: m
ean
(±SD
):H
bA1c
: 7.5
(±0.
8)%
.In
sulin
: 0.6
(±0.
18) U
/kg
BMI:
25.9
(±3.
8) k
g/m
2
Asp
art
part
icip
ants
:N
: 59
End
of s
tudy
: 55
Fem
ale:
61%
Mea
n ag
e (±
SD):
42.3
(±12
) yBa
selin
e va
lues
: mea
n (±
SD):
HbA
1c: 7
.3 (±
0.7)
%.
Insu
lin: 0
.7 (±
0.76
) U/k
gBM
I: 26
.7 (±
3.8)
kg/
m2
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
137
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Mel
ki e
t al
.,19
9811
6
Fran
ce, m
ulti-
cent
re,
rand
omise
dcr
osso
ver
Dur
atio
n: s
olub
le3
mon
ths,
lisp
ro3
mon
ths
Trea
tmen
t 1:
CSI
I+
sol
uble
(Act
rapi
d)
Trea
tmen
t 2:
CSI
I+
lisp
ro(H
umal
og)
Min
imed
506
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
ofou
tcom
es: h
ypos
,te
chni
cal o
r m
etab
olic
inci
dent
s, e
.g.
keto
nuria
, rec
orde
d in
note
book
. Num
ber
ofhy
pos
defin
ed b
y BG
<3.
0m
mol
/l or
ver
ylo
w B
G <
2.0
mm
ol/l
note
d at
mon
thly
visit
s. S
atisf
actio
nqu
estio
nnai
re a
t en
dof
stu
dy. L
ast
30 d
ays
of 1
st p
erio
d an
alys
ed
Allo
catio
n to
tre
atm
ent
grou
ps: r
ando
mise
d af
ter
run-
in. N
ofu
rthe
r de
tails
Run-
in/w
ash-
out
perio
d: 4
wee
k ru
n-in
of C
SII +
reg
ular
prio
r to
ran
dom
isatio
n. A
naly
sis o
f HbA
1cpe
rfor
med
onl
ydu
ring
first
per
iod
due
to a
car
ryov
er e
ffect
. Ana
lysis
of
hypo
s re
stric
ted
to la
st 3
0 da
ys o
f firs
t pe
riod
to a
void
conf
usio
n du
e to
car
ryov
er e
ffect
Com
para
bilit
y of
tre
atm
ent
grou
ps: H
bA1c
repo
rted
for
each
grou
p at
bas
elin
e, n
o fu
rthe
r de
tails
. Pos
sibili
ty o
f bia
s as
seco
nd p
erio
d ig
nore
d fo
r an
alys
is of
hyp
os a
nd H
bA1c
Ana
lysis
: not
ITT
ana
lysis
. Mea
ns (S
E) r
epor
ted.
SD
calc
ulat
ed b
y re
view
er. A
ll te
sts
2-ta
iled.
Com
paris
ons
for
base
line
char
acte
ristic
s us
ed χ
2or
Fish
er’s
exa
ct t
est.
AN
OVA
for
cont
inuo
us v
aria
bles
app
lied
for
cros
sove
r st
udy
with
per
iod,
tre
atm
ent
grou
p an
d in
tera
ctio
n fa
ctor
s. In
cas
eof
sig
nific
ant
inte
ract
ion
or c
arry
over
effe
ct, o
nly
the
first
perio
d w
as u
sed
in c
ompa
rison
s. C
ateg
oric
al d
ata
com
pare
dby
sig
n te
st b
etw
een
perio
ds a
nd b
etw
een
trea
tmen
t gr
oups
.St
udy
cond
ucte
d in
5 c
entr
es, i
nter
cent
re v
aria
bilit
y no
tas
sess
edSa
mpl
e siz
e/po
wer
cal
cula
tion:
not
rep
orte
dG
ener
alisa
bilit
y: a
dults
with
Typ
e 1
diab
etes
who
hav
e us
ed a
pum
p fo
r at
leas
t 1
year
Reas
ons
for
with
draw
al: p
erso
nal r
easo
ns a
nd la
ck o
fpr
otoc
ol c
ompl
ianc
e (n
= 1
)C
onfli
ct o
f int
eres
t: st
udy
supp
orte
d by
Lill
y Fr
ance
Popu
latio
n: a
dults
with
Ty
pe 1
dia
bete
s ag
ed 1
8–60
y, t
reat
ed b
y C
SII w
ithre
gula
r in
sulin
for
at le
ast
1 y
befo
re e
nrol
men
tC
riter
ia: H
bA1c
<8.
5%, n
egat
ive
C-p
eptid
e re
spon
se,
anti-
insu
lin a
ntib
odie
s <
70%
. Non
e ha
d un
trea
ted
retin
opat
hy, i
mpa
ired
rena
l fun
ctio
n, g
astr
icne
urop
athy
, BM
I>30
, dai
ly in
sulin
dos
e >
2IU
/kg,
hist
ory
of h
ypo
unaw
aren
ess
or s
ever
e di
seas
e th
atco
uld
inte
rfer
e w
ith s
tudy
Part
icip
ants
: N
: 39
End
of s
tudy
: 38
Fem
ale:
43.
6%M
ean
age
(±SD
): 39
.4 (±
9.4)
yD
iabe
tes
dura
tion
mea
n (±
SD):
22.5
(±10
) yBa
selin
e va
lues
: mea
n (±
SD):
HbA
1c: 7
.84
(±0.
75)%
(HbA
1cso
lubl
e 7.
74 ±
1.23
; lisp
ro 7
.97
±0.
8)In
sulin
: 0.5
7 (±
0.12
) IU
/day
BMI:
24.4
(±2.
5) k
g/m
2
cont
inue
d
Appendix 9
138
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Rask
in e
t al
.,20
0111
7
USA
,m
ultic
entr
e,ra
ndom
ised
cros
sove
r
Dur
atio
n: li
spro
3m
onth
s, s
olub
le3
mon
ths
Trea
tmen
t 1:
CSI
I+
lisp
ro(H
umal
og)
Trea
tmen
t 2:
CSI
I+
sol
uble
(Vel
osul
in)
Min
imed
504
,50
4-S,
506
or
507
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
ofou
tcom
es: h
ypos
reco
rded
in d
iary
: pt
belie
ved
(or
othe
rpe
rson
obs
erve
d) t
hey
wer
e ex
perie
ncin
g at
leas
t on
e sig
n or
sym
ptom
ass
ocia
ted
with
hyp
o, o
r BG
<3.
0m
mol
/l. H
yper
= 2
or
mor
e BG
>13
.8m
mol
/l in
2–4
hno
t re
spon
ding
to
insu
lin b
olus
es o
rot
herw
ise u
nexp
lain
ed
Allo
catio
n to
tre
atm
ent
grou
ps: r
ando
mise
d af
ter
run-
in.
Stat
es t
hat
‘rand
omisa
tion
was
suc
h th
at t
here
wou
ld b
eap
prox
imat
ely
equa
l num
bers
of p
ts in
eac
h tr
eatm
ent
sequ
ence
at
each
inve
stig
ativ
e sit
e’. N
o fu
rthe
r de
tails
giv
enRu
n-in
/was
h-ou
t pe
riod:
2–4
wee
k ru
n-in
with
reg
ular
insu
lin. N
o w
ash-
out
perio
d. N
o ca
rryo
ver
or in
vest
igat
oref
fect
s di
scov
ered
by
AN
OVA
, the
refo
re d
ata
from
trea
tmen
t pe
riods
com
bine
dC
ompa
rabi
lity
of t
reat
men
t gr
oups
: dat
a pr
esen
ted
for
both
grou
ps. S
imila
r in
pro
port
ion
of fe
mal
es, a
ge, w
eigh
t an
ddu
ratio
n of
dia
bete
s, b
ut s
tatis
tical
ana
lysis
not
pre
sent
ed.
No
signi
fican
t di
ffere
nce
in H
bA1c
Ana
lysis
: sta
tes
that
ITT
ana
lysis
per
form
ed. 2
-tai
led
test
s,sig
nific
ance
leve
l 0.0
5. A
NO
VA u
sed
to e
xam
ine
for
carr
yove
r, in
vest
igat
or o
r tr
eatm
ent
effe
cts.
Mul
ticen
tre:
inve
stig
ator
effe
cts
asse
ssed
by
AN
OVA
and
non
e id
entif
ied
Sam
ple
size/
pow
er c
alcu
latio
n: n
ot r
epor
ted
Gen
eral
isabi
lity:
Typ
e 1
diab
etes
with
acc
epta
ble
com
plia
nce
and
at le
ast
6 m
onth
s w
ith C
SII
Out
com
e m
easu
res:
HbA
1cm
easu
red
at e
nd o
f eac
htr
eatm
ent
perio
d. In
sulin
dos
es u
sed
on d
ay b
efor
e ea
chcl
inic
visi
t (e
very
6 w
eeks
) rec
orde
d in
dia
ryRe
ason
s fo
r w
ithdr
awal
: per
sona
l rea
sons
(n=
1).
Con
flict
of i
nter
est:
Spon
sore
d by
Eli
Lilly
and
Com
pany
Popu
latio
n: T
ype
1 di
abet
es a
ge 1
3–60
y w
ho h
adac
hiev
ed a
ccep
tabl
e co
mpl
ianc
e w
ith in
sulin
pum
psth
erap
y an
d a
nutr
ition
al r
egim
en a
nd h
ad b
een
trea
ted
with
CSI
I for
at
leas
t 6
mon
ths
Excl
usio
n cr
iteria
: tot
al g
lyca
ted
haem
oglo
bin
valu
es o
rH
bA1c
>2.
0 tim
es u
pper
lim
it of
nor
mal
ran
ge,
clin
ical
ly s
igni
fican
t re
nal,
hepa
tic o
r ca
rdia
c di
seas
e,ca
ncer
, dru
g or
alc
ohol
abu
se, i
nsul
in a
llerg
y, r
ecur
rent
seve
re h
ypog
lyca
emia
, ana
emia
, life
exp
ecta
ncy
<3
y,la
ctat
ing,
pre
gnan
t or
inte
ndin
g to
bec
ome
preg
nant
or
requ
ire d
ilutio
n of
insu
lin in
pum
p. 3
pts
disc
ontin
ued
befo
re r
ando
misa
tion
(2 fo
r pe
rson
al r
easo
ns, 1
faile
dto
mee
t cr
iteria
)
N: 5
9En
d of
stu
dy: 5
8.Se
quen
ce li
spro
the
n re
gula
r in
sulin
: En
d of
stu
dy: 2
8Fe
mal
e: 4
2.9%
Mea
n ag
e (±
SD):
40.5
(±8.
7) y
Dia
bete
s du
ratio
n m
ean
(±SD
): 18
.8 (±
7.6)
yBa
selin
e va
lues
: mea
n (±
SD):
HbA
1c: 7
.9 (±
1.1)
%
Wei
ght:
78.3
(±17
.9) k
g.
Sequ
ence
reg
ular
the
n in
sulin
lisp
ro:
End
of s
tudy
: 30
Fem
ale:
53.
3%M
ean
age
(±SD
): 37
.8 (±
9.7)
yD
iabe
tes
dura
tion
mea
n (±
SD):
17.4
(±8.
5) y
Base
line
valu
es: m
ean
(±SD
):H
bA1c
: 7.6
(±0.
8)%
, p=
0.2
34.
Wei
ght:
77.3
(±16
.7) k
g
Prev
ious
reg
imen
: 2–4
wee
ks r
un-in
with
CSI
I +re
gula
r in
sulin
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
139
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Renn
er e
t al
.,19
9911
4
Ger
man
y,m
ultic
entr
e,ra
ndom
ised
cros
sove
r
Dur
atio
n: li
spro
4m
onth
s, s
olub
le4
mon
ths
Trea
tmen
t 1:
CSI
I+
lisp
ro(H
umal
og).
Min
imed
506
Trea
tmen
t 2:
CSI
I+
sol
uble
(Hum
insu
lin).
Dise
tron
ic H
tron
-10
0
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
ofou
tcom
es: h
ypo:
<3.
5m
mol
/l an
d/or
sym
ptom
s, r
ecor
ded
in d
iary
. Dia
bete
sTr
eatm
ent
Satis
fact
ion
Que
stio
nnai
reco
mpl
eted
at
begi
nnin
g an
d en
d of
each
tre
atm
ent
arm
,m
axim
um s
core
48
Allo
catio
n to
tre
atm
ent
grou
ps: r
ando
mise
d af
ter
run-
in. N
ofu
rthe
r de
tails
Run-
in/w
ash-
out
perio
d: 4
wee
ks r
un-in
with
CSI
I + r
egul
arin
sulin
. No
was
h-ou
t. Tr
eatm
ent
orde
r ef
fect
not
rep
orte
dC
ompa
rabi
lity
of t
reat
men
t gr
oups
: not
rep
orte
dA
naly
sis: A
NO
VA w
ith p
ts, t
reat
men
t an
d pe
riod
term
s.Fr
iedm
an’s
ran
k-su
m t
est
if no
t no
rmal
ly d
istrib
uted
(HbA
1c,
patie
nt p
refe
renc
e). F
requ
ency
of h
ypos
and
cat
hete
roc
clus
ions
use
d M
cNem
ar’s
tes
t. In
sulin
dos
e ca
lcul
ated
by
revi
ewer
from
mea
ns o
f 7 r
epor
ted
mea
l-rel
ated
bas
al a
ndbo
lus
valu
es. T
otal
bas
elin
e an
d 4-
mon
th in
sulin
cal
cula
ted
byre
view
erSa
mpl
e siz
e/po
wer
cal
cula
tion:
not
rep
orte
dG
ener
alisa
bilit
y: w
ell-c
ontr
olle
d Ty
pe 1
dia
bete
s w
ith a
t le
ast
6 m
onth
s C
SII u
seO
utco
me
mea
sure
s: D
iabe
tes
Trea
tmen
t Sa
tisfa
ctio
nQ
uest
ionn
aire
rep
orte
d to
be
valid
ated
Oth
er: t
otal
insu
lin d
ose
calc
ulat
ed b
y re
view
er fr
om b
asal
brea
kfas
t, lu
nch
and
dinn
er; b
olus
bre
akfa
st lu
nch
and
dinn
er;
and
snac
ksC
onfli
ct o
f int
eres
t: 2
auth
ors
hold
sto
ck in
and
/or
rece
ived
hono
raria
from
Min
iMed
and
/or
Eli L
illy.
A 3
rd a
utho
rre
ceiv
ed g
rant
and
res
earc
h su
ppor
t fr
om E
li Li
lly
Popu
latio
n: T
ype
1 di
abet
es >
2 y
and
trea
tmen
t w
ithC
SII f
or a
t le
ast
6 m
onth
sEx
clus
ion
crite
ria: k
now
n in
sulin
alle
rgy,
car
diov
ascu
lar
or c
ereb
rova
scul
ar s
ympt
oms
of a
ther
scle
rosis
, can
cer,
rena
l or
hepa
tic fa
ilure
, sig
ns o
f dru
g ab
use,
life
-th
reat
enin
g di
seas
e, p
regn
ant,
lact
atin
g or
pla
nnin
gpr
egna
ncy
Part
icip
ants
: N
: 113
End
of s
tudy
: 113
Fem
ale:
46.
9%M
ean
age
(±SD
): 37
.1 (±
11.6
) yD
iabe
tes
dura
tion
mea
n (±
SD):
19.1
(±9.
2) y
Base
line
valu
es: m
ean
(±SD
):H
bA1c
: 7.2
4 (±
1)%
Insu
lin: 8
3.4
IU/d
ay (c
alcu
late
d by
rev
iew
er)
BMI:
24.7
(±1.
1) k
g/m
2
Prev
ious
reg
imen
: CSI
I, du
ratio
n 5.
3 ±
3.9
y
cont
inue
d
Appendix 9
140
Stud
yIn
terv
enti
ons
Subj
ects
Out
com
e m
easu
res
Met
hodo
logy
Zin
man
et
al.,
1997
118
Can
ada,
sin
gle-
cent
re,
rand
omise
dcr
osso
ver
Dur
atio
n: li
spro
3m
onth
s, s
olub
le3
mon
ths
Trea
tmen
t 1:
CSI
I+
lisp
ro
Trea
tmen
t 2:
CSI
I+
sol
uble
(Hum
ulin
)
Dise
tron
ic H
tron
V100
.
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
ofou
tcom
es: i
nsul
in d
ose
and
hypo
s re
cord
ed in
diar
y. H
bA1c
mea
sure
dm
onth
ly. H
ypo
= B
G<
3m
mol
/l or
sym
ptom
s. B
efor
era
ndom
isatio
nfr
eque
ncy
12.7
(8.8
)ep
isode
s/30
day
s, o
r if
conf
irmed
by
BG 8
.4(7
.1)/
30da
ys. S
ever
ehy
pos
as d
efin
ed b
yD
CC
T
Allo
catio
n to
tre
atm
ent
grou
ps: d
oubl
e-bl
ind.
Ran
dom
ised
afte
r ru
n-in
Run-
in/w
ash-
out
perio
d: 1
6 pt
s on
MD
I sw
itche
d to
CSI
I +re
gula
r 3
mon
ths
befo
re r
ando
misa
tion.
All
pts
rece
ived
CSI
I+
reg
ular
for
1 m
onth
run
-inC
ompa
rabi
lity
of t
reat
men
t gr
oups
: not
rep
orte
dA
naly
sis: A
NO
VA u
sed
mod
els
with
tre
atm
ent
sequ
ence
(stu
dy g
roup
), pe
riod
and
trea
tmen
t te
rms.
Ord
inal
var
iabl
esw
ith W
ilcox
on’s
ran
k-su
m t
est
(tre
atm
ent
sequ
ence
, per
iod
and
trea
tmen
t te
rm e
valu
ated
in t
hree
sep
arat
e an
alys
es).
Cat
egor
ical
var
iabl
es u
sed
χ2 , McN
emar
’s a
nd/o
r Fi
sher
’sex
act
test
sSa
mpl
e siz
e/po
wer
cal
cula
tion:
not
rep
orte
dG
ener
alisa
bilit
y: a
dults
with
Typ
e 1
diab
etes
pre
viou
sly o
nin
tens
ive
insu
lin t
hera
py o
r C
SII
Con
flict
of i
nter
est:
supp
orte
d by
an
unre
stric
ted
rese
arch
gran
t fr
om E
li Li
lly C
anad
a
Popu
latio
n: a
dults
with
Typ
e 1
diab
etes
and
at
leas
t 3
mon
ths
with
CSI
I. N
one
had
signi
fican
t en
doge
nous
insu
lin s
ecre
tion
as a
sses
sed
by t
he m
easu
rem
ent
offa
stin
g C
-pep
tide
(<0.
02 n
mol
/l)Ex
clus
ion
crite
ria: s
ever
e re
tinop
athy
or
neur
opat
hy,
>1
seve
re p
ast
year
. Mild
to
mod
erat
e re
tinop
athy
was
pre
sent
in 1
6 (5
3%),
mild
to
mod
erat
ene
urop
athy
in 1
2 (4
0%)
Part
icip
ants
: N
: 30
End
of s
tudy
: 30
Fem
ale:
56.
7%M
ean
age
(±SD
): 35
.1 (±
8.2)
y, m
in. 2
6, m
ax. 5
1D
iabe
tes
dura
tion
mea
n (±
SD):
17.5
(±8.
8) y
, min
. 6,
max
. 44
Base
line
valu
es: m
ean
(±SD
):H
bA1c
: 8.0
3 (±
0.71
)%
BMI:
24.8
(±2.
7) k
g/m
2
Wei
ght:
72.7
(±9.
9) k
gPr
evio
us r
egim
en: 1
4 on
CSI
I>3
mon
ths.
Inte
nsifi
edin
sulin
the
rapy
with
pen
(7) o
r sy
ringe
(9) b
efor
esw
itchi
ng t
o C
SII +
reg
ular
insu
lin 3
mon
ths
prio
r to
rand
omisa
tion
Schm
auss
et
al.,
1998
115
Ger
man
y, s
ingl
e-ce
ntre
,ra
ndom
ised
cros
sove
r
Dur
atio
n: li
spro
3m
onth
s, s
olub
le3
mon
ths
Trea
tmen
t 1:
CSI
I+
lisp
ro(H
umal
og)
Trea
tmen
t 2:
CSI
I+
sol
uble
(Hum
ulin
)
Dise
tron
ic H
tron
V100
Reas
on fo
r C
SII:
not
repo
rted
Ass
essm
ent
ofou
tcom
es:
hypo
s an
d ad
vers
eev
ents
rec
orde
d at
clin
ic v
isits
.H
ypos
= B
G<
3.5
mm
ol/l
Allo
catio
n to
tre
atm
ent
grou
ps: r
ando
mise
d af
ter
run-
in, n
ofu
rthe
r de
tails
Run-
in/w
ash-
out
perio
d: r
un-in
per
iod
4w
eeks
. No
was
h-ou
tTr
eatm
ent
orde
r ef
fect
: not
disc
usse
dC
ompa
rabi
lity
of t
reat
men
t gr
oups
: HbA
1cat
ran
dom
isatio
ngi
ven
for
each
gro
up, n
o fu
rthe
r de
tails
Ana
lysis
: AN
OVA
with
rep
eate
d m
easu
res
appl
ied
toev
alua
te t
he in
fluen
ce o
f gro
up a
nd t
ime
on c
ontin
uous
varia
bles
. Wilc
oxon
ran
k te
st fo
r hy
pos.
Mea
n (S
EM)
pres
ente
d. S
D c
alcu
late
d by
rev
iew
er. H
bA1c
at e
nd o
fpe
riods
1 a
nd 2
for
each
tre
atm
ent
aver
aged
by
revi
ewer
Sam
ple
size/
pow
er c
alcu
latio
n: n
ot r
epor
ted
Gen
eral
isabi
lity:
adu
lts w
ith w
ell-c
ontr
olle
d Ty
pe 1
dia
bete
s,w
ith a
t le
ast
6 m
onth
s C
SII u
se a
nd n
o se
vere
dia
betic
com
plic
atio
nsO
utco
me
mea
sure
s: H
bA1c
reco
rded
at
end
of r
un-in
,cr
osso
ver
and
end
of s
tudy
Con
flict
of i
nter
est:
supp
orte
d by
an
unre
stric
ted
rese
arch
gran
t fr
om E
li Li
lly, G
erm
any
Popu
latio
n: w
ell-c
ontr
olle
d Ty
pe 1
dia
bete
s, u
se o
fin
tens
ified
insu
lin t
hera
py a
t le
ast
2y,
with
CSI
I for
at
leas
t 6
mon
ths
prio
r to
stu
dy. A
ge 1
8–65
y.
Excl
usio
n cr
iteria
: kno
wn
alle
rgy
to in
sulin
, sev
ere
com
plic
atio
ns o
f dia
bete
s, in
adeq
uate
met
abol
icco
ntro
l (H
bA1c
>10
%),
life-
thre
aten
ing
dise
ase,
dru
gab
use,
wom
en p
regn
ant
or p
lann
ing
preg
nanc
y
Part
icip
ants
: N
: 11
End
of s
tudy
: 11
Fem
ale:
54.
5%M
ean
age
(±SD
): 30
(±8.
3) y
Dia
bete
s du
ratio
n m
ean
(±SD
): 14
(±3.
3) y
Base
line
valu
es: m
ean
(±SD
):Li
spro
HbA
1c: 6
.3 (±
0.7)
%
Solu
ble
HbA
1c: 6
.7 (±
1.3)
%, p
= n
sBM
I: 24
(±3.
3) k
g/m
2
Health Technology Assessment 2004; Vol. 8: No. 43
141
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Appendix 10
Summary of results: adults with Type 1 diabetes
Appendix 10
142
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
eW
eigh
tLi
pids
Pati
ent
pref
eren
ce/q
ualit
y of
life
Adv
erse
eve
nts
Bak
et a
l.,19
8787
Rand
omise
dcr
osso
ver
Dur
atio
n:C
SII
6m
onth
s,M
DI
6m
onth
sN
: 20
End
of s
tudy
:16
CSI
I + s
olub
leM
onth
: 6M
ean:
46.
6U
/day
SD: 1
0
Mon
th: 6
Mea
n: 7
6.6
kgSD
: 9.6
Not
repo
rted
No.
pre
ferr
ing:
not
rep
orte
dFr
eedo
m fr
om in
ject
ions
: 44%
75%
disc
omfo
rted
by
butt
erfly
nee
dle
(on
CSI
I), 5
6% d
iscom
fort
ed b
y in
fuse
rits
elf a
t ho
me,
21%
disc
omfo
rted
by
infu
ser
at w
ork
QoL
: not
rep
orte
d
Mod
erat
e hy
po: 2
.3 (%
of B
Gva
lues
bel
ow 4
mm
ol/l)
Seve
re h
ypo:
1 (r
equi
ring
hosp
ital t
reat
men
t)D
KA
: 0
Gly
cate
d H
b: n
otre
port
edBl
ood
gluc
ose:
Mon
th: 6
Mea
n: 7
.7SD
: 0.7
Bode
et
al.,
1996
81
Non
-ra
ndom
ised
cros
sove
rD
urat
ion:
MD
I min
. of
12 m
onth
s.C
SII m
ean
37m
onth
s.N
: 55
CSI
I + s
olub
leM
onth
: 12
Mea
n: 3
6.4
U/d
aySD
: 12.
1M
onth
: 24
Mea
n: 3
9.6
U/d
aySD
: 14.
4M
onth
: 36
Mea
n: 3
7.7
U/d
aySD
: 13.
1M
onth
: 48
Mea
n: 3
7.8
U/d
aySD
: 14.
2N
ote:
N=
54
(12
mon
ths)
; 38
(24
mon
ths)
; 25
(36
mon
ths)
; 19
(48
mon
ths)
Mon
th: 1
2M
ean:
68.
1kg
SD: 1
4.1
Mon
th: 2
4M
ean:
69.
7kg
SD: 1
5.2
Mon
th: 3
6M
ean:
70.
0kg
SD: 1
4.6
Mon
th: 4
8M
ean:
68
kgSD
: 15.
2N
ote:
N=
55
(12
mon
ths)
; 41
(24
mon
ths)
; 25
(36
mon
ths)
; 19
(48
mon
ths)
Not
repo
rted
No.
pre
ferr
ing:
not
rep
orte
dQ
oL: n
ot r
epor
ted
Seve
re h
ypo:
12
Not
e: h
ypo
even
ts p
er 1
00pa
tient
yea
rs (n
), sig
nific
ance
from
MD
I: 12
mon
ths
22 (5
5),
p<
0.00
01; 2
4m
onth
s 26
(50)
, p<
0.01
; 36
mon
ths
39(3
3), p
<0.
0001
; 48
mon
ths
36 (2
5), p
<0.
01.
Of 2
5 pt
s w
ith H
bA1c
≥8.
0%,
rate
of s
ever
e hy
pos
decl
ined
signi
fican
tly, f
rom
84
even
tspe
r 10
0 pa
tient
yea
rs a
tba
selin
e (M
DI)
to 8
eve
nts
inye
ar 1
(p<
0.00
01).
The
30
pts
with
bas
elin
e H
bA1c
<8.
0% a
lso e
xper
ienc
ed a
decl
ine,
from
183
to
33se
vere
hyp
os p
er 1
00 p
atie
ntye
ars
(p=
0.00
05).
DK
A r
ates
wer
e no
tsig
nific
antly
diff
eren
t be
twee
nM
DI a
nd C
SII p
hase
s (1
4.6
vs7.
2ev
ents
per
100
pat
ient
year
s, r
espe
ctiv
ely)
HbA
1cre
port
edM
onth
: 12
Mea
n: 7
.4%
SD: 1
.2M
onth
: 24
Mea
n: 7
.7%
SD: 1
.7M
onth
: 36
Mea
n: 7
.4%
SD: 1
.7M
onth
: 48
Mea
n: 7
.4%
SD: 1
.2N
ote:
N=
55
(12
mon
ths)
; 41
(24
mon
ths)
; 26
(36
mon
ths)
; 20
(48
mon
ths)
MD
I +re
gula
rM
onth
: 6M
ean:
51.
9 U
/day
SD: 1
2.9
Sign
ifica
nce:
p<
0.0
5
Mon
th: 6
Mea
n: 7
4.8
kgSD
: 9.7
Sign
ifica
nce:
p<
0.0
1
Not
repo
rted
Pref
errin
g: 8
0% w
ante
d to
con
tinue
with
MD
I95
% fo
und
pen
easy
to
hand
le, 3
0%hi
gh fr
eque
ncy
of in
ject
ions
disa
dvan
tage
Of t
otal
: ind
epen
denc
e of
fixe
dm
ealti
mes
an
adva
ntag
e w
ith M
DI a
ndC
SII b
y 85
%Q
oL: n
ot r
epor
ted
Mod
erat
e hy
po: 4
.1 (%
of B
Gva
lues
bel
ow 4
mm
ol/l)
, p
< 0
.02
Seve
re h
ypo:
0D
KA
: 0
Bloo
d G
luco
se:
Mon
th: 6
Mea
n: 7
.9SD
: 0.7
Sign
ifica
nce:
p=
ns
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
143
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
eW
eigh
tLi
pids
Pati
ent
pref
eren
ce/q
ualit
y of
life
Adv
erse
eve
nts
MD
I +re
gula
r M
onth
: 12
Mea
n: 4
2.9
U/d
aySD
: 17.
9Si
gnifi
canc
e: p
= n
s
Mon
th: 1
2M
ean:
67.
4kg
SD: 1
4.4
Sign
ifica
nce:
p=
ns
Not
repo
rted
No.
pre
ferr
ing:
not
rep
orte
d Q
oL: n
ot r
epor
ted
Seve
re h
ypo:
76
MD
I: hy
po e
vent
s pe
r 10
0pa
tient
yea
rs (n
):138
(55)
HbA
1cre
port
ed
Mon
th: 1
2M
ean:
7.7
%SD
: 1.5
Sign
ifica
nce:
p=
ns
Not
e: a
sig
nific
ant
impr
ovem
ent
in m
ean
HbA
1cfr
om b
asel
ine
(MD
I) to
yea
r 1
was
seen
with
pat
ient
s w
hoha
d un
acce
ptab
legl
ycae
mic
con
trol
at
base
line
(MD
I), d
efin
edas
HbA
1c≥
8.0%
(8.9
±0.
8% v
s 8.
1 ±
1.0%
,p
= 0
.000
4). I
n co
ntra
st,
the
30 p
atie
nts
who
had
bett
er g
lyca
emic
con
trol
at b
asel
ine,
def
ined
as
HbA
1c<
8.0%
sho
wed
no s
igni
fican
t di
ffere
nce
in c
ontr
ol fr
om b
asel
ine
to y
ear
1 (6
.7 ±
1.1%
vs 6
.8 ±
1.2%
, p=
ns)
cont
inue
d
Appendix 10
144
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
eW
eigh
tLi
pids
Pati
ent
pref
eren
ce/q
ualit
y of
life
Adv
erse
eve
nts
Brin
chm
ann-
Han
sen
et a
l.,19
8876
,78
RCT
(par
alle
l)D
urat
ion:
CSI
I mea
n40
mon
ths
(24–
60).
MD
I mea
n41
mon
ths
(21–
47)
CSI
I N: 1
5M
DI N
: 15
CSI
I + s
olub
leM
onth
: 24
Mea
n: 0
.68
U/k
gSD
: 0.1
9 (S
EM 0
.05)
Mon
th: 2
4M
ean:
70.
5kg
SD: 8
.9 (S
EM 2
.3)
Not
repo
rted
No.
pre
ferr
ing:
not
rep
orte
dQ
oL: n
ot r
epor
ted
Hyp
ogly
caem
ia m
ean
(SEM
)va
lues
: (%
of h
ome
BG<
2.5
mm
ol/l)
: at
rand
omisa
tion
= 6
(2),
afte
r2
year
s =
11
(2),
p<
0.05
.Sy
mpt
omat
icep
isode
s/w
eek/
patie
nt: a
tra
ndom
isatio
n =
2.3
(0.6
); at
2ye
ars
= 1
.7 (0
.3)
Hyp
o co
ma:
2 (2
) D
KA
: 2Su
bcut
aneo
us a
bsce
ss: 8
(6)
Retin
opat
hy =
impr
oved
(14%
), un
chan
ged
(29%
),w
orse
ned
(57%
) (on
e pa
tient
had
an a
llerg
y to
fluo
resc
ein)
HbA
1re
port
ed
Mon
th: 3
Mea
n: 8
.9%
Mon
th: 6
Mea
n: 9
.2%
Mon
th: 1
2M
ean:
8.5
%M
onth
: 24
Mea
n: 8
.7%
SD: 1
.16
Mon
th: 4
1 M
ean:
9.1
%
MD
I +re
gula
rM
onth
: 24
Mea
n: 0
.72
U/k
gSD
: 0.1
2 (S
EM 0
.03)
Sign
ifica
nce:
p=
ns
Mon
th: 2
4M
ean:
75.
1kg
SD: 1
1.2
(SEM
2.9
)Si
gnifi
canc
e: p
= n
s
Not
repo
rted
No.
pre
ferr
ing:
not
rep
orte
dQ
oL: n
ot r
epor
ted
Hyp
ogly
caem
ia m
ean
(SEM
)va
lues
: (%
of h
ome
BG<
2.5
mm
ol/l)
: at
rand
omisa
tion
= 8
(3),
afte
r2
year
s =
7 (1
), p
= n
s.Sy
mpt
omat
icep
isode
s/w
eek/
patie
nt: a
tra
ndom
isatio
n =
1.5
(0.3
); at
2ye
ars
= 1
.2 (0
.2).
Hyp
o co
ma:
14
(6) v
s C
SII,
p<
0.00
1.D
KA
: 0 (v
s C
SII,
p=
ns)
Subc
utan
eous
abs
cess
: 0 v
sC
SII,
p <
0.0
1Re
tinop
athy
= im
prov
ed(7
%),
unch
ange
d (4
3%),
wor
sene
d (5
0%) (
one
patie
ntha
d an
alle
rgy
to fl
uore
scei
n)
HbA
1re
port
ed
Mon
th: 3
Mea
n: 8
.7%
Mon
th: 6
Mea
n: 8
.8%
Mon
th: 1
2M
ean:
8.5
%M
onth
: 24
Mea
n: 9
.1%
SD: 1
.55
Mon
th: 4
1M
ean:
9.4
%
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
145
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
eW
eigh
tLi
pids
Pati
ent
pref
eren
ce/q
ualit
y of
life
Adv
erse
eve
nts
Chi
asso
n,
et a
l., 1
98482
Non
-ra
ndom
ised
cros
sove
rD
urat
ion:
CSI
I3
mon
ths,
MD
I3
mon
ths
N=
12
CSI
I + s
olub
leM
onth
: 3M
ean:
43.
9 U
/day
SD: 1
0
Not
rep
orte
dN
otre
port
edN
o. p
refe
rrin
g: 7
(58%
)Q
oL: n
ot r
epor
ted
Not
rep
orte
dH
bA1
repo
rted
M
onth
: 3M
ean:
9.1
%SD
: 1.0
4
Haa
kens
et
al.,
199
083
Non
-ra
ndom
ised
cros
sove
rD
urat
ion:
CSI
I6
mon
ths,
MD
I 6m
onth
sN
: 52
End
of s
tudy
:35
CSI
I + s
olub
leM
onth
: 6M
ean:
0.6
4 U
/kg/
day
SD: 0
.18
CSI
I sig
nific
antly
low
erth
an d
urin
g C
T(p
<0.
001)
,M
DI/h
uman
ultr
alen
te(p
<0.
001)
,M
DI/h
uman
isop
hane
(p<
0.00
5)
Mon
th: 6
Mea
n: 6
9.1
kgSD
: 8.6
3St
art
of C
SII v
s en
d of
CSI
I (st
art
of M
DI),
p
= n
s
Not
repo
rted
23/5
2 (4
4%) p
refe
rred
CSI
I: Re
ason
s:C
ontr
olFl
exib
ility
in t
erm
s of
food
inta
keal
low
edA
void
freq
uent
inje
ctio
ns: 2
(9%
)G
reat
er s
ense
of w
ell-b
eing
: 2 (9
%)
QoL
: not
rep
orte
d
Inje
ctio
n sit
e in
fect
ion:
1D
KA
: 1Se
vere
hyp
os(e
piso
des/
patie
nt/m
onth
) =1/
7.5;
req
uirin
g gl
ucos
ei.v
./glu
cago
n =
1/9
6,as
sista
nce
but
not
requ
iring
gluc
ose
i.v./g
luca
gon
= 1
/8.
No.
of m
ild s
ubje
ctiv
ehy
pos/
wee
k: 1
.4 (S
D 0
.92)
HbA
1re
port
ed
Mon
th: 6
Mea
n: 9
.6%
SD: 2
.47
MD
I +re
gula
rM
onth
: 6M
DI/h
uman
isop
hane
Mea
n: 0
.72
SD: 0
.25
Sign
ifica
nce:
p<
0.0
05M
I/hum
an u
ltral
ente
Mea
n: 0
.79
SD: 0
.25
Sign
ifica
nce:
p>
0.0
01(is
opha
ne v
sul
tral
ente
, p<
0.0
1)
Mon
th: 6
Mea
n: 7
0.6
kgSD
: 8.6
3Si
gnifi
canc
e: p
< 0
.05
Star
t of
CSI
I vs
end
ofM
DI,
p <
0.0
2
Not
repo
rted
20/5
2 (3
8%) p
refe
rred
MD
I. Re
ason
s:C
ontr
ol: 1
7M
ore
flexi
ble
lifes
tyle
tha
n C
T w
ithou
tne
ed t
o w
ear
a de
vice
Pref
erre
d hu
man
isop
hane
at
bedt
ime:
16 (4
6%)
Pref
erre
d hu
man
ultr
alen
te a
t be
dtim
e:7
(20%
)N
o pr
efer
ence
bet
wee
n be
dtim
ein
sulin
: 11
(31%
)Q
oL: n
ot r
epor
ted
Inje
ctio
n sit
e in
fect
ion:
0D
KA
: 0Se
vere
hyp
os(e
piso
des/
patie
nt/m
onth
) =1/
5; r
equi
ring
gluc
ose
i.v./g
luca
gon
= 1
/79,
assis
tanc
e bu
t no
t re
quiri
nggl
ucos
e i.v
./glu
cago
n =
1/5
.N
o. o
f mild
sub
ject
ive
hypo
s/w
eek:
isop
hane
1.4
(SD
0.92
), ul
tral
ente
1.5
(SD
1.1
7)
HbA
1re
port
ed
Mon
th: 6
Mea
n: 9
.8%
SD: 1
.85
Sign
ifica
nce:
p=
ns
cont
inue
d
MD
I +re
gula
rM
onth
: 3M
ean:
56.
1 U
/day
SD: 2
0.4
Sign
ifica
nce:
p<
0.0
1
Not
rep
orte
dN
otre
port
edN
o. p
refe
rrin
g: 5
(42%
)Q
oL: n
ot r
epor
ted
Not
rep
orte
dH
bA1
repo
rted
M
onth
: 3M
ean:
8.7
%SD
: 1.3
9Si
gnifi
canc
e: p
= n
s
Appendix 10
146
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
eW
eigh
tLi
pids
Pati
ent
pref
eren
ce/q
ualit
y of
life
Adv
erse
eve
nts
Han
aire
-Br
outin
et
al.,
2000
80
Rand
omise
dcr
osso
ver
Dur
atio
n:C
SII
4m
onth
s,M
DI
4m
onth
sN
: 41
End
of s
tudy
:40
CSI
I + li
spro
Mon
th: 4
Mea
n: 3
8.5
U/d
aySD
: 9.8
Not
e: m
ean
valu
eov
er 4
mon
ths
Mon
th: 4
Mea
n: 6
8.7
kgSD
: 10
Not
repo
rted
No.
pre
ferr
ing:
29
(72.
5%)
Of t
hese
29,
21
prev
ious
ly o
n C
SII +
regu
lar
and
8 on
MD
IQ
oL: n
ot r
epor
ted
Seve
re h
ypo:
3 e
vent
s, 2
pts
Not
e: h
ypos
dur
ing
last
14da
ys o
f eac
h tr
eatm
ent
perio
d: 3
.9 (4
.2).
Seve
re h
ypos
did
not
res
ult
inco
ma
or s
eizu
res,
ext
erna
lhe
lp n
eede
d to
tak
e su
gar,
but
gluc
agon
or
gluc
ose
inje
ctio
nno
t re
quire
d
HbA
1cre
port
ed
Mon
th: 4
Mea
n: 7
.89%
SD: 0
.77
Not
e: m
ean
valu
e ov
er4
mon
ths
Hom
e et
al.,
1982
88
Rand
omise
dcr
osso
ver
Dur
atio
n:C
SII
2.5
mon
ths,
MD
I 2.5
mon
ths
N: 1
1En
d of
stu
dy:
10
CSI
I + s
olub
leM
onth
: 2.5
Mea
n: 5
1 U
/day
SD: 1
5.8
Not
e: In
divi
dual
chan
ges
+21
% t
o–4
3% o
f tw
ice-
daily
inje
ctio
n do
se.
Requ
irem
ents
rem
aine
d st
eady
from
2nd
wee
k
Mon
th: 2
.5M
ean:
68.
9kg
SD: 9
.2N
ote:
wei
ght
gain
clin
ical
ly s
igni
fican
tpr
oble
m o
n C
SII,
onoc
casio
n lim
iting
adju
stm
ent
of in
sulin
dosa
ge
Cho
lest
erol
:en
d of
stu
dyM
ean:
5SD
: 0.6
No.
pre
ferr
ing:
8 (8
0%) p
refe
rred
pum
p to
MD
I, 4
wish
ed t
o co
ntin
uew
ith it
. Uni
vers
ally
reg
arde
d as
too
larg
e an
d un
com
fort
able
QoL
: not
rep
orte
d
Pum
p m
alfu
nctio
n: 1
(1 p
t ×
3ep
isode
s in
16
h du
e to
faul
typu
mp)
Seve
re h
ypo:
0.6
(0–3
)ep
isode
s pe
r pt
biw
eekl
y[B
asel
ine
hypo
1.1
(0–4
)ep
isode
s pe
r pt
biw
eekl
y]1
pt r
equi
red
intr
aven
ous
gluc
ose
HbA
1re
port
ed
Mon
th: 2
.5M
ean:
10%
SD: 2
.2N
ote:
5 r
each
ed n
orm
al(<
8.2%
) at
som
e po
int
durin
g pu
mps
MD
I +re
gula
rM
onth
: 2.5
Mea
n: 8
0 U
/day
SD: 2
8.5
Sign
ifica
nce:
p=
0.0
04N
ote:
diff
eren
ce C
SII
– M
DI d
ue t
odi
ffere
nce
in b
asal
insu
lin r
equi
rem
ents
[CSI
I 29
(3) U
/day
vs
MD
I 53
(9) U
/day
,p
=0.
013]
. Ind
ivid
ual
incr
ease
s co
mpa
red
with
pum
p: +
20%
to
+12
0%, t
wic
e da
ily:
+14
% t
o +
74%
Mon
th: 2
.5M
ean:
67
kgSD
: 9.2
Sign
ifica
nce:
p=
0.0
23
Cho
lest
erol
:En
d of
stu
dyM
ean:
5.3
SD: 0
.6
No.
pre
ferr
ing:
0N
one
wish
ed t
o co
ntin
ue w
ith 3
×in
ject
ions
due
to
inco
nven
ienc
eQ
oL: n
ot r
epor
ted
Seve
re h
ypo:
0.8
(0–7
)ep
isode
s pe
r pt
biw
eekl
yH
bA1
repo
rted
M
onth
: 2.5
Mea
n: 1
1.7%
SD: 1
.9Si
gnifi
canc
e: p
= 0
.026
Not
e: 1
pt
reac
hed
norm
al d
urin
g M
DI
MD
I + li
spro
Mon
th: 4
Mea
n: 4
7.3
U/d
aySD
: 14.
9Si
gnifi
canc
e: p
< 0
.000
1N
ote:
mea
n nu
mbe
r of
NPH
inje
ctio
ns =
2.65
day
at e
nd o
f MD
Ipe
riod
Mon
th: 4
Mea
n: 6
9kg
SD: 9
.5Si
gnifi
canc
e: p
= n
s
Not
repo
rted
No.
pre
ferr
ing:
11
(27.
5%)
Of t
hese
11,
10
prev
ious
ly o
n C
SII a
nd1
MD
IQ
oL: n
ot r
epor
ted
Seve
re h
ypo:
1 e
vent
, 1 p
tN
ote:
hyp
os d
urin
g la
st14
days
of e
ach
trea
tmen
tpe
riod
= 4
.3 (3
.9),
p=
ns.
Seve
re h
ypos
did
not
res
ult
inco
ma
or s
eizu
res,
ext
erna
lhe
lp n
eede
d to
tak
e su
gar,
but
gluc
agon
or
gluc
ose
inje
ctio
nno
t re
quire
d
HbA
1cre
port
ed
Mon
th: 4
Mea
n: 8
.24%
SD: 0
.77
Sign
ifica
nce:
p<
0.0
01N
ote:
mea
n va
lue
over
4m
onth
s
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
147
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
eW
eigh
tLi
pids
Pati
ent
pref
eren
ce/q
ualit
y of
life
Adv
erse
eve
nts
Nat
han
and
Lou,
198
289
Rand
omise
dcr
osso
ver
Dur
atio
n:Ea
ch p
erio
d8–
12 w
eeks
N: 5
CSI
I + s
olub
leM
onth
: 3M
ean:
35.
4 U
/day
SD: 1
1.5
Not
rep
orte
dN
otre
port
edN
o. p
refe
rrin
g: n
ot r
epor
ted
Not
e: e
ase
of a
dmin
ister
ing
and
adju
stin
g in
sulin
dos
es –
not
ed. A
decr
ease
in h
ypos
and
impr
oved
feel
ing
of w
ell-b
eing
con
trib
uted
to
acce
ptan
ce o
f pum
p; d
raw
back
s =
diffi
culty
for
wom
en in
dre
ssin
g to
acco
mm
odat
e pu
mp
and
rem
oval
of
pum
p fo
r w
ater
spo
rts
QoL
: not
rep
orte
d
Dur
ing
2.5
patie
nt-y
ears
CSI
I=
sub
cuta
neou
s ab
sces
sne
edin
g or
al a
ntib
iotic
s (1
),pu
mp
mal
func
tions
(2)
cath
eter
disl
odge
d at
nig
ht (6
)H
ypo
reac
tions
n/w
eek
at3
mon
ths
= 1
(0.2
5)
HbA
1cre
port
ed
Mon
th: 3
Mea
n: 5
.4%
SD: 0
.34
Nos
adin
i et
al.,
198
885
RCT
(par
alle
l)D
urat
ion:
12m
onth
sC
SII f
ixed
basa
l rat
e(F
BR):
19M
DI:
15C
SII h
ighe
rda
wn
insu
lin(H
OR)
: 10
CSI
I + s
olub
le(F
BR)
Not
rep
orte
dN
ot r
epor
ted
Not
repo
rted
No.
pre
ferr
ing:
not
rep
orte
dQ
oL: n
ot r
epor
ted
Mea
n ev
ents
/pt/
y:H
yper
18
(5)
Keto
tic e
vent
s 0.
13 (0
.02)
Mild
hyp
o 36
(10)
Seve
re h
ypo
0.14
(0.0
5)Su
dden
dea
th 0
.04
(1pt
dea
din
bed
). M
alfu
nctio
n of
pum
pca
used
abo
ut 2
0% o
f ket
otic
episo
des
in b
oth
CSI
I gro
ups.
23%
(FBR
) and
27%
(HO
R)ca
used
by
infe
ctiv
e di
seas
e
HbA
1cre
port
ed
Mon
th: 1
2M
ean:
6.3
%SD
: 0.7
MD
I +re
gula
rN
ot r
epor
ted
Not
rep
orte
dN
otre
port
edN
o. p
refe
rrin
g: n
ot r
epor
ted
QoL
: not
rep
orte
dH
yper
20
(3),
p=
ns.
Ke
totic
eve
nts
0.03
(0.0
1),
p<
0.01
Mild
hyp
o 59
(12)
, p<
0.01
Seve
re h
ypo
0.42
(0.1
5),
p<
0.01
(all
vs C
SII–
FBR)
HbA
1cre
port
ed
Mon
th: 1
2M
ean:
7.1
%SD
: 0.9
Sign
ifica
nce:
p<
0.0
1M
DI v
s C
SII–
FBR
and
CSI
I–H
OR,
p<
0.0
1
MD
I +re
gula
rM
onth
: 3M
ean:
48.
8 U
/day
SD: 1
3.8
Not
rep
orte
dN
otre
port
edN
o. p
refe
rrin
g: n
ot r
epor
ted
QoL
: not
rep
orte
dH
ypo
reac
tions
n/w
eek
at3
mon
ths
= 2
.5 (0
.61)
HbA
1cre
port
ed
Mon
th: 3
Mea
n: 7
.88%
SD: 1
.37
Sign
ifica
nce:
pno
tre
port
edN
ote:
n=
4
cont
inue
d
Appendix 10
148
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
eW
eigh
tLi
pids
Pati
ent
pref
eren
ce/q
ualit
y of
life
Adv
erse
eve
nts
CSI
I + s
olub
le(H
OR)
Not
rep
orte
dN
ot r
epor
ted
Not
repo
rted
No.
pre
ferr
ing:
not
rep
orte
dQ
oL: n
ot r
epor
ted
Hyp
er 1
7 (4
) vs
MD
I, p
= n
sKe
totic
eve
nts
0.16
(0.0
3) v
sM
DI,
p<
0.0
5M
ild h
ypo
30 (1
1) v
s M
DI,
p<
0.0
1Se
vere
hyp
o 0.
16 (0
.09)
vs
MD
I, p
< 0
.01
CSI
I–H
OR
vs C
SII-F
BR, a
llp
=ns
HbA
1cre
port
ed
Mon
th: 1
2M
ean:
6.1
%SD
: 0.9
Not
e: v
s M
DI p
< 0
.01
Saur
brey
etal
., 19
8890
Rand
omise
dcr
osso
ver
Dur
atio
n:C
SII
2.5
mon
ths,
MD
I2.
5m
onth
sN
: 21
End
of s
tudy
:19
CSI
I + s
olub
leC
olle
cted
dur
ing
trea
tmen
t pe
riod
Mea
n: 0
.62
U/d
aySD
: 0.1
7
Not
rep
orte
dN
otre
port
edN
o. p
refe
rrin
g: 6
(31.
5%)
Reas
ons
Life
styl
e: 4
2Q
oL: n
ot r
epor
ted
Seve
re h
ypo:
0N
o. o
f sub
ject
ive
orbi
oche
mic
al (B
G <
2.5
mm
ol/l)
hypo
s in
eac
h gr
oup,
p=
ns
Keto
acid
osis
= 1
(due
to
acut
e ga
stro
ente
ritis
and
failu
re t
o ta
ke e
xtra
insu
lin,
not
pum
p m
alfu
nctio
n)Su
bcut
aneo
us in
fect
ion
(3tim
es) i
n sa
me
patie
nt, o
nere
quire
d su
rgic
al in
cisio
n
HbA
1cre
port
edM
onth
: 2.5
Mea
n: 7
.5%
MD
I +re
gula
rC
olle
cted
dur
ing
stud
yM
ean:
0.6
4 U
/day
SD: 0
.13
Not
rep
orte
dN
otre
port
edN
o. p
refe
rrin
g: 1
2 (6
3%)
Reas
ons:
Life
styl
e: 5
81
(5%
) wer
e un
sure
of p
refe
renc
e(e
ither
ICT
or
CSI
I)Q
oL: n
ot r
epor
ted
Seve
re h
ypo:
0H
bA1c
repo
rted
M
onth
: 2.5
Mea
n: 7
.5%
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
149
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
eW
eigh
tLi
pids
Pati
ent
pref
eren
ce/q
ualit
y of
life
Adv
erse
eve
nts
Schi
ffrin
and
Belm
onte
,19
8291
Rand
omise
dcr
osso
ver
Dur
atio
n:C
SII
6m
onth
s,M
DI
6m
onth
s.N
: 20
End
of s
tudy
:16
CSI
I + s
olub
leM
onth
: 6M
ean:
42
U/d
ayN
ot r
epor
ted
Cho
lest
erol
:En
d of
stu
dyM
ean:
163
mg/
dlSD
: 15
Base
line:
190
(9) m
g/dl
Trig
lyce
ride:
end
of s
tudy
Mea
n:82
mg/
dl
SD: 1
5
No.
pre
ferr
ing:
7/1
6 (4
4%) o
f pts
chos
e C
SII f
or e
asie
r gl
ycae
mic
con
trol
and
also
gre
ater
flex
ibili
ty fo
r m
eal
hour
s an
d m
eal s
ize
CSI
I rej
ecte
d fo
r re
ason
s of
bul
k,di
scom
fort
, a fe
elin
g of
dep
ende
ncy
onth
e m
achi
ne &
dist
ortio
n of
bod
yim
age
70%
wou
ld r
ecom
men
d C
SII t
o a
diab
etic
frie
ndQ
oL: n
ot r
epor
ted
Mild
hyp
o: 1
86M
oder
ate
hypo
: 10
Seve
re h
ypo:
1(s
ever
e hy
po in
1 p
t du
e to
dela
ying
mea
l afte
r bo
lus
insu
lin t
aken
– in
trav
enou
sgl
ucos
e w
as a
dmin
ister
ed)
HbA
1re
port
ed
Mon
th: 3
Mea
n: 8
.1%
SD: 0
.6M
onth
: 6M
ean:
8.2
%SD
: 0.5
Not
e: H
bA1
repo
rted
mon
thly.
Dat
a fo
rm
onth
s 3
and
6ex
trac
ted
only
MD
I +re
gula
rM
onth
: 6M
ean:
44
U/d
aySi
gnifi
canc
e: p
= n
s
Not
rep
orte
dC
hole
ster
ol:
end
of s
tudy
Mea
n: 1
62m
g/dl
SD: 1
2Ba
selin
e: 1
83(8
)mg/
dl.
Mea
nba
selin
ech
oles
tero
lfo
r w
hole
grou
p: 1
86.5
(70–
200)
.Ba
selin
eva
lue
vs C
SII
and
MD
I at
end
of s
tudy
,p
<0.
01Tr
igly
cerid
e:en
d of
stu
dyM
ean:
81m
g/dl
SD
: 12,
p
= n
s
No.
pre
ferr
ing:
7 (4
4%)
30%
wou
ld r
ecom
men
d M
DI t
o a
diab
etic
frie
nd.
(2 p
ts r
ever
ted
to c
onve
ntio
nal)
QoL
: not
rep
orte
d
Mild
hyp
o: 1
89, p
= n
sM
oder
ate
hypo
: 11
p=
ns
Seve
re h
ypo:
0, p
= n
s
HbA
1re
port
ed
Mon
th: 3
Mea
n: 8
.4%
SD: 0
.7Si
gnifi
canc
e: p
= n
sM
onth
: 6M
ean:
8.4
%SD
: 0.5
Sign
ifica
nce:
p=
ns
cont
inue
d
Appendix 10
150
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
eW
eigh
tLi
pids
Pati
ent
pref
eren
ce/q
ualit
y of
life
Adv
erse
eve
nts
Schm
itz
et a
l., 1
98986
Rand
omise
dcr
osso
ver
Dur
atio
n:C
SII
6m
onth
s,M
DI
6m
onth
sN
= 1
0
CSI
I + s
olub
leM
onth
: 6M
ean:
0.5
5 U
/kg/
day
SD: 0
.1Ba
selin
e vs
CSI
I,p
=0.
03
Mon
th: 6
Mea
n: 7
3kg
SD: 1
7.8
Base
line
vs C
SII,
p=
0.01
Not
repo
rted
No.
pre
ferr
ing:
4 (4
0%)
QoL
: not
rep
orte
dSe
vere
hyp
o: 0
DK
A: 0
HbA
1cre
port
ed
Mon
th: 6
Mea
n: 7
%SD
: 1Ba
selin
e vs
CSI
I, p
= 0
.03
Tsui
et
al.,
2001
79
RCT
(par
alle
l)D
urat
ion:
9m
onth
s.C
SII:
13En
d of
stu
dy:
12 MD
I: 14
End
of s
tudy
:14
CSI
I + li
spro
Mon
th: 9
Mea
n: 0
.6 U
/kg
SD: 0
.2
Not
rep
orte
dN
otre
port
edN
o. p
refe
rrin
g: n
ot r
epor
ted
QoL
sco
re:
Satis
fact
ion
75.6
Impa
ct: 6
9.9
Dia
betic
wor
ry: 8
5.2
Soci
al w
orry
: 89.
6G
loba
l Hea
lth :
68.2
Resp
onse
rat
e: 8
5%
Seve
re h
ypo:
6M
ean
num
ber
of h
ypos
eac
hm
onth
, 8.9
(3 m
onth
s), 7
.2(6
mon
ths)
, 7.0
(9 m
onth
s),
8.0
(ove
rall)
, all
p=
ns
HbA
1cre
port
edM
onth
: 3M
ean:
6.9
2%M
onth
: 6M
ean:
7.1
9%M
onth
: 9M
ean:
7.3
8%Po
oled
est
imat
es o
fpa
ralle
l and
cro
ssov
erst
udy:
diff
eren
ce C
SII
and
MD
I = 0
.09%
(95%
CI –
0.09
to
+0.
26),
p>
0.1
.Pr
obab
ility
tha
t tr
uetr
eatm
ent
effe
ct is
mor
eex
trem
e th
an ±
0.5%
is<
0.00
001
MD
I +re
gula
rM
onth
: 6M
ean:
0.6
U/k
g/da
ySD
: 0.1
1Si
gnifi
canc
e: p
= 0
.02
Not
e: b
asel
ine
vs M
DI,
p=
ns
Mon
th: 6
Mea
n: 7
2.8
kgSD
: 18.
4Si
gnifi
canc
e: p
= n
sBa
selin
e vs
MD
I,p
=ns
Not
repo
rted
No.
pre
ferr
ing:
6 (6
0%)
QoL
: not
rep
orte
dSe
vere
hyp
o: 0
DK
A: 0
HbA
1cre
port
ed
Mon
th: 6
Mea
n: 7
.7%
SD: 1
Sign
ifica
nce:
p=
0.0
02Ba
selin
e vs
MD
I, p
= n
s
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
151
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
eW
eigh
tLi
pids
Pati
ent
pref
eren
ce/q
ualit
y of
life
Adv
erse
eve
nts
MD
I + li
spro
Mon
th: 9
Mea
n: 0
.7 U
/kg
SD: 0
.2Tr
eatm
ent
effe
ct(a
djus
ted
for
base
line)
:–0
.10
(85%
CI –
0.26
to 0
.07)
, p>
0.1
0Si
gnifi
canc
e gi
ven
for
diffe
rent
CSI
I – M
DI
–0.1
0
Not
rep
orte
dN
otre
port
edN
o. p
refe
rrin
g: n
ot r
epor
ted
QoL
: Sa
tisfa
ctio
n: 6
8.3
Impa
ct: 6
8.4
Dia
betic
wor
ry: 7
9.8
Soci
al w
orry
: 94.
0G
loba
l Hea
lth :
67.3
p=
ns
for
any
of t
he s
ubsc
ales
Resp
onse
rat
e: 1
00%
Seve
re h
ypo:
4, p
> 0
.10
Mea
n nu
mbe
r of
hyp
os e
ach
mon
th, 5
.0 (3
mon
ths)
, 9.0
(6m
onth
s), 9
.2 (9
mon
ths)
,7.
4 (o
vera
ll), a
ll p
= n
s vs
CSI
I Re
lativ
e tr
eatm
ent
effe
ct(C
SII–
MD
I)/M
DI f
or o
vera
llno
. of h
ypos
= +
9% (9
5% C
I–3
7 to
+87
) p>
0.1
0
HbA
1cre
port
ed
Mon
th: 3
Mea
n: 7
.55%
Trea
tmen
t ef
fect
(adj
uste
d fo
r ba
selin
e):
–0.2
1 (9
5% C
I –0.
59 t
o0.
17),
p>
0.1
0M
onth
: 6M
ean:
7.6
2%Tr
eatm
ent
effe
ct(a
djus
ted
for
base
line)
:–0
.01
(95%
CI –
0.44
to
0.42
), p
> 0
.10
Mon
th: 9
Mea
n: 7
.56%
Trea
tmen
t ef
fect
(adj
uste
d fo
r ba
selin
e):
0.25
(95%
CI –
0.19
to
0.68
), p
> 0
.10
Not
e: t
reat
men
t ef
fect
CSI
I – M
DI s
how
nco
rrec
ted
for
base
line
diffe
renc
es
cont
inue
d
Appendix 10
152
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
eW
eigh
tLi
pids
Pati
ent
pref
eren
ce/q
ualit
y of
life
Adv
erse
eve
nts
Zie
gler
et
al.,
1990
84
RCT
(par
alle
l)D
urat
ion:
24m
onth
sC
SII:
49En
d of
stu
dy:
36 MD
I: 47
End
of s
tudy
:37
CSI
I + s
olub
leN
o va
lues
giv
en, b
utp
=ns
dur
ing
follo
w-
up
No
valu
es g
iven
, but
note
d in
disc
ussio
nth
at a
vera
ge w
eigh
tga
in a
fter
2 y
=1.
3kg
. BM
I at
mon
ths
6–24
= 2
2.5
No
valu
esgi
ven,
but
note
d in
disc
ussio
nth
atch
oles
tero
lle
vels
incr
ease
dw
ithin
the
norm
al r
ange
durin
g th
est
udy
with
out
diffe
renc
esbe
twee
ntr
eatm
ent
grou
ps
No.
pre
ferr
ing:
not
rep
orte
d4
pts
chan
ged
to M
DI d
urin
g st
udy
due
to t
echn
ical
pro
blem
s w
ith p
ump
QoL
: not
rep
orte
d
No.
of m
etab
olic
com
plic
atio
ns/6
mon
ths
mea
n(r
ange
). Re
view
er h
asca
lcul
ated
mea
n of
the
4
×6-
mon
th p
erio
ds t
o gi
veto
tals
belo
wM
ild h
ypo
(BG
<50
mg/
dl):
15.5
(0–7
0)M
ild h
ypo
(sym
ptom
s): 1
9.8
(0–1
04)
Seve
re h
ypo
(req
uirin
gas
sista
nce)
: 0.0
91 (0
–7)
Seve
re h
ypo
(com
a): 0
.101
(0–4
)Ke
tosis
(BG
>20
0m
g/dl
and
keto
nuria
): 6.
5 (0
–57)
Keto
acid
osis:
0.0
46 (0
–1)
Freq
uenc
y of
pts
(no.
of
even
ts/1
00 y
)H
ypo
requ
iring
ass
istan
ce:
13.9
(18.
1)H
ypo
com
a: 3
0.6
(19.
4)D
KA
: 13.
9 (9
.7) [
true
the
rapy
at t
ime
of e
vent
: 15.
4 (1
4.1)
]
HbA
1re
port
ed
Mon
th: 6
Mea
n: 8
.2%
Mon
th: 1
2M
ean:
8.5
%M
onth
: 18
Mea
n: 8
.5%
Mon
th: 2
4M
ean:
8.6
MD
I +re
gula
rN
o va
lues
giv
en, b
ut
p=
ns
durin
g fo
llow
-up
As
for
CSI
I, av
erag
ew
eigh
t ga
in a
fter
2y
= 2
.8kg
. BM
I at
mon
ths
6–24
= 2
4.2,
p<
0.05
No.
pre
ferr
ing:
not
rep
orte
d7
pts
chan
ged
trea
tmen
t gr
oup
durin
gst
udy
due
to e
xplic
it re
ques
t to
hav
e a
pum
pQ
oL: N
ot r
epor
ted
Mild
hyp
o (B
G <
50m
g/dl
):10
.6 (0
–53)
Mild
hyp
o (s
ympt
oms)
: 10.
1(0
–81)
Seve
re h
ypo
(req
uirin
gas
sista
nce)
: 0.0
70 (0
–2)
Seve
re h
ypo
(com
a): 0
.048
(0–1
)Ke
tosis
(BG
>20
0 m
g/dl
and
keto
nuria
): 7.
8 (0
–85)
Keto
acid
osis:
0.0
40 (0
–3)
Freq
uenc
y of
pts
(no.
of
even
ts/1
00 y
)H
ypo
requ
iring
ass
istan
ce:
16.2
(13.
5)H
ypo
com
a: 1
3.5
(9.5
)D
KA
: 8.1
(8.1
) [tr
ue t
hera
pyat
tim
e of
eve
nt: 5
.9 (2
.9)]
HbA
1re
port
ed
Mon
th: 6
Mea
n: 8
.8%
Sign
ifica
nce:
p<
0.0
5M
onth
: 12
Mea
n: 8
.7%
Sign
ifica
nce:
p=
ns
Mon
th: 1
8M
ean:
8.4
%Si
gnifi
canc
e: p
= n
sM
onth
: 24
Mea
n: 8
.5Si
gnifi
canc
e: p
= n
s
Health Technology Assessment 2004; Vol. 8: No. 43
153
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Appendix 11
Summary of results: pregnancy
Appendix 11
154
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
eP
regn
ancy
out
com
esPa
tien
t pr
efer
ence
Adv
erse
eve
nts
Burk
art
etal
., 19
8895
RCT
(par
alle
l)
Dur
atio
n:9
mon
ths
CSI
I: 48
MD
I: 41
CSI
I +so
lubl
eN
ot r
epor
ted
Mea
n du
ratio
n: 3
8.4
wee
ks
Live
birt
hs: 4
7/48
(97.
9%)
Mea
n bi
rth
wei
ght:
3082
.7 (u
nits
not
spec
ified
)
No.
Cae
sare
an: 2
4/48
Com
plic
atio
ns in
pre
gnan
cy:
Pyel
onep
hriti
s: 5
Prem
atur
e la
bour
: 4
Prem
atur
e ru
ptur
e: 2
Prem
atur
e de
liver
y: 7
Pr
e-ec
lam
psia
: 3G
row
th r
etar
datio
n: 3
Hyp
ogly
caem
ia: 5
St
ill b
irth:
1
Keto
acid
osis:
1To
tal:
12/4
8 pr
egna
ncie
s ha
d on
e or
mor
e co
mpl
icat
ions
Inci
denc
e of
com
plic
atio
ns in
bot
hgr
oups
is li
nked
to
the
seve
rity
ofm
ater
nal d
iabe
tes
(Whi
te’s
crit
eria
).Bi
rthw
eigh
t de
crea
ses
with
sev
erity
.D
iffer
ence
s in
mea
n bi
rthw
eigh
ts, p
=ns
Feta
l out
com
e:H
ealth
y ne
wbo
rns:
40
Min
or s
ympt
oms:
8
Maj
or s
ympt
oms:
0I.v
. glu
cose
: 14
Ove
rall
mor
talit
y up
to
age
1 y
= 2
/48
No.
pre
ferr
ing:
not
repo
rted
QoL
: not
rep
orte
d
See
com
plic
atio
ns in
preg
nanc
yN
ot r
epor
ted.
Sta
tes
that
patie
nts
sele
cted
in t
heC
SII o
r M
DI g
roup
had
norm
al m
ean
gluc
ose
and
HbA
1<
7.5%
at
leas
t fr
omth
e en
d of
the
firs
ttr
imes
ter
on
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
155
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
eP
regn
ancy
out
com
esPa
tien
t pr
efer
ence
Adv
erse
eve
nts
CSI
I +re
gula
rN
ot r
epor
ted
Mea
n du
ratio
n: 3
8.2
wee
ks
Live
birt
hs: 1
00%
Mea
n bi
rth
wei
ght:
3319
.5 (u
nits
not
spec
ified
)
No.
Cae
sare
an: 1
5/41
(p =
ns)
Com
plic
atio
ns in
pre
gnan
cy:
Pyel
onep
hriti
s: 6
Prem
atur
e la
bour
: 3Pr
emat
ure
rupt
ure:
1Pr
emat
ure
deliv
ery:
4Pr
e-ec
lam
psia
: 3
Gro
wth
ret
arda
tion:
1H
ypog
lyca
emia
: 3St
ill b
irth:
0Ke
toac
idos
is: 0
Tota
l: 13
/41
preg
nanc
ies
had
one
orm
ore
com
plic
atio
ns (v
s C
SII,
p=
ns)
Feta
l out
com
e:H
ealth
y ne
wbo
rns:
34
Min
or s
ympt
oms:
5M
ajor
sym
ptom
s: 2
I.v
. glu
cose
9 (a
ll p
= n
s vs
CSI
I)O
vera
ll m
orta
lity
up t
o ag
e 1
y =
3/4
1,vs
CSI
I p<
0.0
5. (N
ote:
the
p-v
alue
repo
rted
by
the
pape
r ap
pear
s to
be
inco
rrec
t, an
d sh
ould
be
insig
nific
ant)
No.
pre
ferr
ing:
not
repo
rted
QoL
: not
rep
orte
d
See
com
plic
atio
ns in
preg
nanc
y
cont
inue
d
Appendix 11
156
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
eP
regn
ancy
out
com
esPa
tien
t pr
efer
ence
Adv
erse
eve
nts
Car
ta e
t al
.,19
8694
RCT
(par
alle
l)
Dur
atio
n:9
mon
ths
Type
1di
abet
es:
CSI
I: 8
MD
I:7
CSI
I +so
lubl
eM
onth
: 3M
ean:
37
U/d
ay
Mon
th: 6
Mea
n: 4
2 U
/day
Mon
th: 9
Mea
n: 4
9.1
U/d
ay
Mea
n du
ratio
n: 2
66.7
day
sSD
: 7.6
Live
birt
hs: 1
00%
Mea
n bi
rth
wei
ght:
3395
gSD
: 407
No.
Cae
sare
an: 3
(37.
5%)
1 pr
eter
m d
ue t
o hy
dram
nios
and
feta
lm
egal
osom
nia.
1 la
rge
for
gest
atio
nal
age.
1 m
acro
som
nic
(>40
00g)
.[A
rms
not
spec
ified
: 2 in
fant
s de
pres
sed,
5 m
etab
olic
mor
bidi
ty (h
ypog
lyca
emia
,hy
poca
lcae
mia
, hyp
erbi
lirub
inae
mia
), 2
cong
enita
l car
diac
mal
form
atio
ns]
No.
pre
ferr
ing:
not
repo
rted
QoL
: not
rep
orte
d
Inje
ctio
n sit
e in
fect
ion:
0Pu
mp
mal
func
tion:
0C
athe
ter
disc
onne
ctio
n: 3
In a
ll pt
s: n
o hy
po c
oma.
1se
vere
hyp
o (g
roup
not
stat
ed).
Mild
hyp
osoc
curr
ed o
n av
erag
e on
ceev
ery
15–2
0 da
ys
Mon
th: 3
Mea
n: 7
.8%
Mon
th: 5
Mea
n: 7
%
Mon
th: 8
Mea
n: 7
%
Mon
th: 9
Mea
n: 6
.5%
MD
I +re
gula
rM
onth
: 3M
ean:
32
U/d
aySi
gnifi
canc
e: p
= n
s
Mon
th: 6
Mea
n: 4
7 U
/day
Sign
ifica
nce:
p=
ns
Mon
th: 9
Mea
n: 6
0.4
U/d
aySi
gnifi
canc
e: p
= n
s
Mea
n du
ratio
n: 2
63.6
day
sSD
: 17.
7
Live
birt
hs: 1
00%
Mea
n bi
rth
wei
ght:
2906
gSD
: 553
Sign
ifica
nce:
p=
ns
No.
Cae
sare
an: 3
(42.
9%)
1 pr
eter
m d
ue t
o m
etro
ttha
gia
caus
edby
cen
tral
pla
cent
a pr
evia
No.
pre
ferr
ing:
not
repo
rted
QoL
: not
rep
orte
d
Mon
th: 3
Mea
n: 8
.5%
Sign
ifica
nce:
p=
ns
Mon
th: 5
Mea
n: 7
.75%
Sign
ifica
nce:
p=
ns
Mon
th: 8
Mea
n: 6
.95%
Sign
ifica
nce:
p=
ns
Mon
th: 9
Mea
n: 6
.45%
Sign
ifica
nce:
p=
ns
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
157
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
eP
regn
ancy
out
com
esPa
tien
t pr
efer
ence
Adv
erse
eve
nts
Car
ta e
t al
.,19
8694
RCT
(par
alle
l)
Dur
atio
n:9
mon
ths
Type
2di
abet
es:
CSI
I: 6
MD
I: 8
CSI
I +so
lubl
eM
onth
: 3M
ean:
23.
1U
/day
Mon
th: 6
Mea
n: 3
6U
/day
Mon
th: 9
Mea
n: 5
2.8
U/d
ay
Mea
n du
ratio
n: 2
72 d
ays
SD: 9
Live
birt
hs: 1
00%
Mea
n bi
rth
wei
ght:
3292
gSD
: 578
No.
Cae
sare
an: 1
(16.
7%)
1 pr
eter
m d
ue t
o pr
emat
ure
rupt
ure
ofm
embr
anes
. 1 la
rge
for
gest
atio
nal a
ge
No.
pre
ferr
ing:
not
repo
rted
QoL
: not
rep
orte
d
Episo
des
of m
ild h
ypos
wer
e fe
w a
nd d
id n
otoc
cur
in a
ll pt
s
Mon
th: 3
Mea
n: 7
%
Mon
th: 5
Mea
n: 6
.75%
Mon
th: 8
Mea
n: 6
.5%
Mon
th: 9
Mea
n: 6
%
MD
I +re
gula
rM
onth
: 3M
ean:
28
U/d
aySi
gnifi
canc
e: p
= n
s
Mon
th: 6
Mea
n: 4
9 U
/day
Sign
ifica
nce:
p=
ns
Mon
th: 9
Mea
n: 5
1.2
U/d
aySi
gnifi
canc
e: p
= n
s
Mea
n du
ratio
n: 2
69 d
ays
SD: 1
1.7
Live
birt
hs: 1
00%
Mea
n bi
rth
wei
ght:
2994
gSD
: 512
Sign
ifica
nce:
p=
ns
No.
Cae
sare
an: 3
(37.
5%)
1 pr
eter
m d
ue t
o pr
emat
ure
rupt
ure
ofm
embr
anes
. 1 s
mal
l for
ges
tatio
nal a
ge
No.
pre
ferr
ing:
not
repo
rted
QoL
: not
rep
orte
d
Mon
th: 3
Mea
n: 8
.1%
Sign
ifica
nce:
p=
ns
Mon
th: 5
Mea
n: 7
.2%
Sign
ifica
nce:
p=
ns
Mon
th: 8
Mea
n: 6
.7%
Sign
ifica
nce:
p=
ns
Mon
th: 9
Mea
n: 6
.7%
Sign
ifica
nce:
p=
ns
cont
inue
d
Appendix 11
158
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
eP
regn
ancy
out
com
esPa
tien
t pr
efer
ence
Adv
erse
eve
nts
Cou
ston
et
al.,
198
693
RCT
(par
alle
l)
Dur
atio
n:9
mon
ths
CSI
I: 11
MD
I: 11
CSI
I + li
spro
Mon
th: 3
Mea
n: 0
.71
U/k
g/da
ySD
: 0.1
6
Mon
th: 6
Mea
n: 1
.02
U/k
g/da
ySD
: 0.5
3
Mon
th: 9
Mea
n: 1
.26
U/k
g/da
ySD
: 0.4
9
Mea
n du
ratio
n: 3
8.1
wee
ksSD
: 2.1
Live
birt
hs: 1
00%
Mea
n bi
rth
wei
ght:
3050
gSD
: 675
No.
Cae
sare
an: 7
(63.
6%)
1bi
rthw
eigh
t ab
ove
90th
per
cent
ile,
2sm
all f
or g
esta
tiona
l age
bab
ies.
Neo
nata
l hyp
ogly
caem
ia in
1
No.
pre
ferr
ing:
not
repo
rted
QoL
: not
rep
orte
d
Inje
ctio
n sit
e in
fect
ion:
0Pu
mp
mal
func
tion:
0M
ild h
ypo:
11
pts
Mod
erat
e hy
po: 6
pts
Seve
re h
ypo:
3 p
ts(8
episo
des)
DK
A: 0
1 hy
perg
lyca
emia
and
keto
nuria
(2da
ys) d
urin
gvi
ral i
nfec
tion.
Cat
hete
r le
akag
e or
occl
usio
n oc
curr
edin
freq
uent
ly a
nd r
esol
ved
quic
kly
Mon
th: 2
Mea
n: 7
%M
onth
: 9M
ean:
6.3
%SD
: 0.6
Not
e: S
igni
fican
cere
duct
ion
from
bas
elin
e(p
<0.
05)
MD
I + li
spro
Mon
th: 3
Mea
n: 1
.01
U//k
g/da
ySD
: 0.2
8Si
gnifi
canc
e: p
= 0
.101
Mon
th: 6
Mea
n: 1
.4 U
//kg/
day
SD: 0
.4Si
gnifi
canc
e: p
= 0
.027
Mon
th: 9
Mea
n: 1
.63
U//k
g/da
ySD
: 0.5
1Si
gnifi
canc
e: p
= 0
.041
Mea
n du
ratio
n: 3
8.8
wee
ksSD
: 1.4
Live
birt
hs: 1
00%
Mea
n bi
rth
wei
ght:
3324
gSD
: 475
Sign
ifica
nce:
p=
ns
No.
Cae
sare
an: 7
(63.
6%)
1 ba
by m
oder
ate
resp
irato
ry d
istre
sssy
ndro
me.
Neo
nata
l hyp
ogly
caem
ia in
1
No.
pre
ferr
ing:
not
repo
rted
QoL
: not
rep
orte
d
Mild
hyp
o: 1
1 pt
sM
oder
ate
hypo
: 10
pts
Seve
re h
ypo:
5 p
ts (2
3ep
isode
s)D
KA
: 0
1 su
bjec
t ha
d 17
hyp
osre
late
d to
die
tary
irreg
ular
ities
Mon
th: 2
Mea
n: 7
.5%
Sign
ifica
nce:
p=
ns
Mon
th: 9
Mea
n: 6
.4%
SD: 0
.4Si
gnifi
canc
e: p
= n
s
Not
e: s
igni
fican
cere
duct
ion
from
bas
elin
e(p
< 0
.05)
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
159
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
eP
regn
ancy
out
com
esPa
tien
t pr
efer
ence
Adv
erse
eve
nts
Nos
ari e
t al
.,19
9392
RCT
(par
alle
l)
Dur
atio
n:9
mon
ths
CSI
I: 16
MD
I: 16
CSI
I +so
lubl
eM
onth
: 3M
ean:
39.
6 U
/day
SD: 2
2.8
Mon
th: 6
Mea
n: 4
8.1
U/d
aySD
: 20.
8
Mon
th: 9
Mea
n: 5
7.3
U/d
aySD
: 21.
6
Not
e: d
ata
at 1
st, 2
nd, 3
rdtr
imes
ter
Mea
n du
ratio
n: 3
8.9
wee
ksSD
: 7.6
Live
birt
hs: 8
7.5%
Mea
n bi
rth
wei
ght:
3130
gSD
: 148
0
No.
Cae
sare
an: 9
(56.
3%)
2 in
trau
terin
e de
aths
. Ges
tatio
nal a
ge a
tde
liver
y (p
regn
ancy
dur
atio
n) p
> 0
.5,
birt
hwei
ght
p>
0.5
. Fet
al c
onge
nita
lm
alfo
rmat
ions
, mac
roso
mia
, low
Apg
arsc
ores
(<7
at 5
min
utes
) not
obs
erve
d.1
CSI
I lar
ge fo
r ag
e. R
espi
rato
ry d
istre
ssin
1 C
SII i
nfan
t. N
eona
tal h
ypo
(pla
sma
gluc
ose
<30
mg/
dl) i
n 1
CSI
I.
No.
pre
ferr
ing:
not
repo
rted
QoL
: not
rep
orte
d
Seve
re h
ypo:
3D
KA
: 3M
onth
: 3M
ean:
6%
SD: 3
.6
Mon
th: 6
Mea
n: 6
.8%
SD: 5
.6
Mon
th: 9
Mea
n: 6
.3%
SD: 2
Not
e: d
ata
at 1
st, 2
nd, 3
rdtr
imes
ter
MD
I +re
gula
rM
onth
: 3M
ean:
36
U/d
aySD
: 24.
4Si
gnifi
canc
e: p
> 0
.5
Mon
th: 6
Mea
n: 4
3.7
U/d
aySD
: 35.
6Si
gnifi
canc
e: p
> 0
.5
Mon
th: 9
Mea
n: 5
4.7
U/d
aySD
: 39.
2Si
gnifi
canc
e: p
> 0
.5
Mea
n du
ratio
n: 3
8.2
wee
ksSD
: 10
Live
birt
hs: 9
3.8%
Mea
n bi
rth
wei
ght:
3010
gSD
: 172
8
No.
Cae
sare
an: 7
(43.
8%)
1 in
trau
terin
e de
ath.
2 M
DI s
mal
l for
age.
1 p
rem
atur
e M
DI b
irth.
Neo
nata
lhy
po (p
lasm
a gl
ucos
e <
30m
g/dl
) in
1M
DI
No.
pre
ferr
ing:
not
repo
rted
QoL
: not
rep
orte
d
Seve
re h
ypo:
1, C
SII v
sM
DI,
p>
0.5
DK
A: 0
, CSI
I vs
MD
I,p
>0.
1
Mon
th: 3
Mea
n: 6
.2%
SD: 1
.6Si
gnifi
canc
e: p
>0.
5
Mon
th: 6
Mea
n: 6
.1%
SD: 2
.4Si
gnifi
canc
e: p
> 0
.5
Mon
th: 9
Mea
n: 6
.2%
SD: 0
.8Si
gnifi
canc
e: p
> 0
.5
Health Technology Assessment 2004; Vol. 8: No. 43
161
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Appendix 12
Summary of results: adolescents
Appendix 12
162
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
ePa
tien
t pr
efer
ence
Adv
erse
eve
nts
Schi
ffrin
et
al.,
198
497
Rand
omise
dcr
osso
ver
Dur
atio
n:C
SII
4m
onth
s,M
DI
4m
onth
s,C
SII+
MD
I4
mon
ths
N: 2
4
CSI
I +so
lubl
eM
onth
: 4M
ean:
44
U/d
aySD
: 12
Not
e: t
otal
insu
lin/d
ay d
urin
gtr
eatm
ent
– no
tim
e po
int
give
nC
SII v
s IC
T a
nd b
asel
ine,
p<
0.0
01
No.
pre
ferr
ing:
11
100%
wou
ldre
com
men
d C
SII t
o a
frie
nd.
QoL
: not
rep
orte
d
Seve
re h
ypo:
1
No
diffe
rent
in h
ypos
from
oth
ertr
eatm
ents
. Mos
tly m
ild a
nd a
vera
ged
one
per
wee
k pe
r pt
. Sam
e pt
s ha
dse
vere
hyp
o on
CSI
I and
ICT
– t
hela
tter
due
to
inad
vert
ent
adm
inist
ratio
n of
ext
ra N
PH d
ose
Mon
th: 3
Mea
n: 8
.75%
M
onth
: 4M
ean:
8.8
%N
ote:
red
uctio
n is
signi
fican
t fr
om b
asel
ine
(p<
0.0
01)
CSI
I vs
CSI
I-IC
T, p
< 0
.05
CSI
I vs
ICT,
p<
0.0
5
MD
I +re
gula
rM
onth
: 4M
ean:
60
U/d
aySD
: 16
Sign
ifica
nce:
p<
0.0
01
Not
e: t
otal
insu
lin/d
ay d
urin
gtr
eatm
ent
– no
tim
e po
int
give
n
No.
pre
ferr
ing:
466
% w
ould
rec
omm
end
ICT
to
a fr
iend
QoL
: not
rep
orte
d
Seve
re h
ypo:
1 (d
ue t
o in
adve
rten
tad
min
istra
tion
of e
xtra
NPH
dos
e)M
onth
: 3M
ean:
9.5
%Si
gnifi
cant
: p<
0.0
5M
onth
: 4M
ean:
9.6
%Si
gnifi
canc
e: p
< 0
.05
Not
e: r
educ
tion
is sig
nific
ant
from
bas
elin
e(p
<0.
001)
CSI
Iov
erni
ght
with
MD
Idu
ring
day
Mon
th: 4
Mea
n: 4
8U
/day
SD: 1
6N
ote:
dur
ing
who
le s
tudy
CSI
I–IC
T v
s IC
T a
nd b
asel
ine,
p
< 0
.001
No.
pre
ferr
ing:
370
% w
ould
rec
omm
end
CSI
I–IC
T t
o a
frie
nd(2
of 2
0 co
mpl
etin
gst
udy
reve
rted
to
twic
eda
ily in
ject
ions
)
QoL
: not
rep
orte
d
Mon
th: 3
Mea
n: 9
%
Mon
th: 4
Mea
n: 9
.3%
Not
e: r
educ
tion
signi
fican
t fr
om b
asel
ine
(p<
0.0
01)
CSI
I vs
CSI
I–IC
T, p
< 0
.05
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
163
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
ePa
tien
t pr
efer
ence
Adv
erse
eve
nts
Tam
borla
neet
al.,
198
996
Rand
omise
dcr
osso
ver
Dur
atio
n:C
SII
6m
onth
s,M
DI
6m
onth
s
N: 1
0
CSI
I +re
gula
rM
onth
: 6M
ean:
1.3
U/d
aySD
: 3.1
6
No.
pre
ferr
ing:
not
repo
rted
QoL
: not
rep
orte
d
Episo
des
of m
ild h
ypos
wer
e co
mm
ondu
ring
each
tre
atm
ent.
Stat
es t
hat
ther
e w
as o
nly
one
seve
re h
ypo
requ
iring
ass
istan
ce d
urin
g ea
ch
6-m
onth
tre
atm
ent
perio
d
Mon
th: 6
Mea
n: 8
.5%
SD: 1
.26
MD
I +re
gula
rM
onth
: 6M
ean:
1.4
U/d
aySD
: 0.3
2Si
gnifi
canc
e: p
=ns
Mon
th: 0
Mea
n: 0
U/d
ay
Insu
lin d
ose
(U/d
ay) w
as in
crea
sed
durin
g bo
th M
DI a
nd C
SII
com
pare
d w
ith b
asel
ine,
p<
0.0
5
No.
pre
ferr
ing:
not
repo
rted
QoL
: not
rep
orte
d
Mon
th: 6
Mea
n: 8
.7%
SD: 1
.58
Sign
ifica
nce:
p=
ns
HbA
1sig
nific
antly
red
uced
com
pare
d w
ithba
selin
e du
ring
both
CSI
I and
MD
I, p
< 0
.05
Health Technology Assessment 2004; Vol. 8: No. 43
165
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Appendix 13
Summary of results: analogue versus soluble insulin
Appendix 13
166
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
eW
eigh
tPa
tien
t pr
efer
ence
Adv
erse
eve
nts
Bode
et
al.,
2002
113
RCT
(par
alle
l)
Dur
atio
n:4
mon
ths
Lisp
ro: 2
8(e
nd o
fst
udy
27)
Regu
lar:
59
(end
of
stud
y 50
)A
spar
t: 59
(end
of
stud
y 55
)
CSI
I + li
spro
Not
rep
orte
dN
ot r
epor
ted
Not
rep
orte
dSe
vere
hyp
o: 0
DK
A: 0
Hyp
ergl
ycae
mia
(>35
0m
g/dl
): 10
≤3
clog
s or
blo
ckag
es: 6
4%H
ypos
[epi
sode
s, r
ate
(SD
)]:
All
repo
rted
: 872
, 10.
5 (8
.1),
vs a
spar
t p
= 0
.044
Hyp
o BG
<50
mg/
dl: 3
59, 4
.4 (4
.7),
vs a
spar
tp
=0.
841
Noc
turn
al h
ypo
BG <
50m
g/dl
: 64,
0.6
(0.6
1), v
sas
part
p=
0.18
9
Mon
th: 4
Mea
n ch
ange
from
base
line:
0.1
8 (0
.84)
%
CSI
I +bu
ffere
dre
gula
r
Not
rep
orte
dN
ot r
epor
ted
Not
rep
orte
dSe
vere
hyp
o: 1
(1.7
%)
DK
A: 0
Hyp
ergl
ycae
mia
(>35
0m
g/dl
): 24
≤3
clog
s or
blo
ckag
es: 7
8%H
ypos
[epi
sode
s, r
ate
(SD
)]:
All
repo
rted
: 166
3, 1
0.5
(8.9
), vs
asp
art
p=
0.0
34H
ypo
BG<
50m
g/dl
: 770
, 4.8
(4.2
), vs
asp
art
p=
0.17
5.N
octu
rnal
hyp
o BG
<50
mg/
dl: 2
07, 0
.9 (0
.97)
, vs
aspa
rt p
=0.
004
Mon
th: 4
Mea
n ch
ange
from
base
line:
0.1
5 (0
.63)
%
CSI
I + a
spar
tN
ot r
epor
ted
Not
rep
orte
dN
ot r
epor
ted
Seve
re h
ypo:
0D
KA
: 0H
yper
glyc
aem
ia (>
350
mg/
dl):
16 (p
=ns
bet
wee
nal
l tre
atm
ents
)
≤3
clog
s or
blo
ckag
es: 7
5%H
erpe
s zo
ster
= 1
Hyp
os [e
piso
des,
rat
e (S
D)]
:A
ll re
port
ed: 1
126,
6.7
(5.4
)H
ypo
BG<
50m
g/dl
: 610
, 3.7
(3.6
)N
octu
rnal
hyp
o BG
<50
mg/
dl: 9
6, 0
.5 (0
.83)
Mon
th: 4
Mea
n ch
ange
from
base
line:
0.0
0 (0
.15)
%
p=
ns
betw
een
grou
ps
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
167
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
eW
eigh
tPa
tien
t pr
efer
ence
Adv
erse
eve
nts
Mel
ki e
t al
.,19
9811
6
Rand
omise
dcr
osso
ver
Dur
atio
n:So
lubl
e3
mon
ths,
lispr
o3
mon
ths
N: 3
9En
d of
stud
y: 3
8
CSI
I +so
lubl
eM
onth
: 3M
ean:
0.5
3 U
/kg/
day
SD: 0
.12
Mon
th: 3
Mea
n: 0
.04
kgSD
: 1.7
9
No.
pre
ferr
ing:
36
(94.
7%)
Reas
ons:
Con
trol
: 89.
4%Li
fest
yle:
84.
2%
% o
f fav
oura
ble
resp
onse
s fo
rlis
pro
varie
d fr
om 8
4 to
92%
acco
rdin
g to
the
que
stio
ns
QoL
: not
rep
orte
d
Cat
hete
r ob
stru
ctio
ns :
9 Se
vere
hyp
o: 3
(7.9
%)
DK
A: 0
Insu
lin p
reci
pita
tion
in c
athe
ter:
1
BG<
3.0
mm
ol/l:
7.0
3 (5
.79)
per
mon
thBG
<2.
0 m
mol
/l: 0
.05
(0.3
1) p
er m
onth
Mon
th: 3
Mea
n: 7
.11%
SD: 0
.92
Not
e: r
educ
tion
from
base
line:
0.6
2 (0
.8)
CSI
I + li
spro
Mon
th: 3
Mea
n: 0
.55
U/k
g/da
ySD
: 0.1
2
Sign
ifica
nce:
p=
ns
Mon
th: 3
Mea
n: 0
.48
kgSD
: 1.6
Sign
ifica
nce:
p
= n
s
No.
pre
ferr
ing:
2 (5
.3%
)
Reas
ons
Con
trol
: 5.3
%Li
fest
yle:
5.3
%
QoL
: not
rep
orte
d
Cat
hete
r ob
stru
ctio
ns: 9
Seve
re h
ypo:
4 (1
0.5%
) pts
, 7 e
vent
sD
KA
: 0In
sulin
pre
cipi
tatio
n in
cat
hete
r: 4
BG<
3.0
mm
ol/l:
7.9
4 (5
.42)
per
mon
th, p
=ns
BG<
2.0
mm
ol/l:
0.4
7 (1
.17)
per
mon
th, p
<0.
050
resu
lted
in c
oma
or s
eizu
res
Mon
th: 3
Mea
n: 7
.88%
SD: 0
.99
Not
e: r
educ
tion
from
base
line:
0.0
9 (0
.92)
Lisp
ro v
s re
gula
r, p
= 0
.01
CSI
I + li
spro
Basa
l 0.3
4 U
/kg,
bolu
s 0.
28 U
/kg
Mon
th: 3
Mea
n: 7
9.2
kgSD
: 17.
1
Not
e: 1
kgga
ined
at
end
of p
erio
d 1,
vs
base
line
p=
ns
Not
rep
orte
dSe
vere
hyp
o: 3
(5.1
%) p
ts (3
epi
sode
s)H
yper
due
to
occl
usio
n: 8
pts
(16
episo
des)
Keto
sis: 1
Hyp
o as
def
ined
: 7 p
ts (8
epi
sode
s)
Mon
th: 3
Mea
n: 7
.41%
SD: 0
.97
Not
e: c
hang
e fr
omba
selin
e: 0
.34
(0.5
9)
CSI
I +so
lubl
eBa
sal 0
.35
U/k
g,bo
lus
0.30
U/k
gM
onth
: 3M
ean:
78.
8kg
SD: 1
7.3
Sign
ifica
nce:
p=
0.78
Not
e: 1
kgga
ined
at
end
of p
erio
d 1,
vs
base
line:
p=
ns
No.
pre
ferr
ing:
not
rep
orte
d
QoL
: not
rep
orte
d
Seve
re h
ypo:
2 (3
.4%
) pts
(3 e
piso
des)
DK
A: 0
Not
e: h
yper
due
to
occl
usio
n: 1
2 pt
s (2
3 ep
isode
s)H
ypo
as d
efin
ed: 7
pts
(11
episo
des)
1 ho
spita
lised
with
feve
r, vo
miti
ng, d
ehyd
ratio
n
(38
pts
repo
rted
109
epi
sode
s of
hyp
er: 3
9 ca
used
by o
cclu
sion,
47
caus
ed b
y ot
her
reas
ons,
23
noid
entif
iabl
e ca
use.
Gro
ups
not
spec
ified
)
Mon
th: 3
Mea
n: 7
.65%
SD: 0
.85
Sign
ifica
nce:
p=
0.0
04
Not
e: c
hang
e fr
omba
selin
e: 0
.09
(0.6
3),
lispr
o vs
reg
ular
p=
0.00
4
cont
inue
d
Rask
in e
t al
.,20
0111
7
Rand
omise
dcr
osso
ver
Dur
atio
n:lis
pro
3m
onth
s,so
lubl
e3
mon
ths
N:5
9 En
d of
stud
y: 5
8
Appendix 13
168
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
eW
eigh
tPa
tien
t pr
efer
ence
Adv
erse
eve
nts
Renn
er
et a
l.,19
9911
4
Rand
omise
dcr
osso
ver
Dur
atio
n:lis
pro
4m
onth
s,so
lubl
e4
mon
ths
N: 1
13
CSI
I + li
spro
Mon
th: 4
Mea
n: 8
3.1
U/d
ayN
ot r
epor
ted
No.
pre
ferr
ing:
not
rep
orte
d
DT
SQ p
atie
nt s
atisf
actio
n (m
ax.
scor
e of
48
poss
ible
):35
.16
(4.2
5)
QoL
: not
rep
orte
d
Inje
ctio
n sit
e in
fect
ion:
4Se
rious
adv
erse
eve
nt n
ot r
elat
ed t
o dr
ug: 1
Hyp
o: m
ean
12.4
(13.
9), m
edia
n 8,
epi
sode
s/pt
.In
fect
ion
(col
d): 1
9.4%
Rhin
itis:
15.
8%C
athe
ter
occl
usio
n: 2
0 (4
2 ep
isode
s)Ke
tosis
: 5
Mon
th: 4
Mea
n: 6
.77%
SD: 0
.88
Schm
auss
et
al.,
1998
115
Rand
omise
dcr
osso
ver
Dur
atio
n:lis
pro
3m
onth
s,so
lubl
e3
mon
ths
N: 1
1
CSI
I + li
spro
Basa
l 19
(6.6
) IU
/day
Bolu
s 1.
4 (0
.3)
IU/1
2g
carb
ohyd
rate
Not
rep
orte
dN
o. p
refe
rrin
g: 1
1 (1
00%
)G
reat
er fl
exib
ility
with
lisp
ro
No
signi
fican
t di
ffere
nce
intr
eatm
ent
satis
fact
ion
note
d
QoL
: not
rep
orte
d
Seve
re h
ypo:
0D
KA
: 0H
ypos
per
30
days
: 4 (2
.98)
No
seve
re a
dver
se e
vent
s re
gist
ered
by
eith
ertr
eatm
ent
Mon
th: 3
Mea
n: 6
%SD
: 0.9
9
CSI
I +so
lubl
eBa
sal 2
0 (3
.3) I
U/d
ay,
p=
nsBo
lus
1.5
(0.3
)IU
/12
gca
rboh
ydra
te, p
=ns
Not
rep
orte
dN
o. p
refe
rrin
g: 0
QoL
: not
rep
orte
d
Seve
re h
ypo:
0D
KA
: 0H
ypos
per
30
days
: 3.2
(2.3
), p
=ns
Mon
th: 3
Mea
n: 6
.35%
SD: 0
.83
Sign
ifica
nce:
p=
ns
Not
e: H
bA1c
at e
nd o
fea
ch t
reat
men
t pe
riod
not
com
bine
d in
pap
er
CSI
I +so
lubl
eM
onth
: 4M
ean:
85.
4 U
/day
Sign
ifica
nce:
p=
ns
Not
rep
orte
dN
o. p
refe
rrin
g: n
ot r
epor
ted
DT
SQ p
atie
nt s
atisf
actio
n (m
ax.
scor
e of
48
poss
ible
): 32
.36
(5.8
7), v
s lis
pro
p<
0.0
01
QoL
: not
rep
orte
d
Inje
ctio
n sit
e in
fect
ion:
2Se
rious
adv
erse
eve
nt n
ot r
elat
ed t
o dr
ug :
6H
ypo:
mea
n 11
.0 (1
1.2)
, med
ian
8, e
piso
de/p
t,p
=ns
.In
fect
ion
(col
d): 2
1.1%
Rhin
itis:
13.
8%C
athe
ter
occl
usio
n: 2
1 (4
5 ep
isode
s), p
= n
sKe
tosis
: 4
Mon
th: 4
Mea
n: 6
.9%
SD: 0
.97
Sign
ifica
nce:
p<
0.0
2
cont
inue
d
Health Technology Assessment 2004; Vol. 8: No. 43
169
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Stud
yTr
eatm
ent
Gly
cate
d H
bIn
sulin
dos
eW
eigh
tPa
tien
t pr
efer
ence
Adv
erse
eve
nts
Zin
man
et
al.,
1997
118
Rand
omise
dcr
osso
ver
Dur
atio
n:lis
pro
3m
onth
s,so
lubl
e3
mon
ths
N: 3
0
CSI
I + li
spro
Mon
th: 3
Mea
n: 4
0.4
U/d
aySD
: 9.3
Mon
th: 3
Mea
n: 7
2.6
kgSD
: 9.9
No.
pre
ferr
ing:
not
rep
orte
d
Qua
lity
of li
fe: n
ot r
epor
ted
Skin
rea
ctio
n: 1
6Se
vere
hyp
o: 0
Disc
olou
ratio
n of
insu
lin in
res
ervo
ir or
cat
hete
r (d
idno
t le
ad t
o ob
stru
ctio
n or
sig
nific
ant
hype
rgly
caem
ia):
2
Hyp
os: 8
.6 (7
.7)/
30 d
ays
(vs
base
line
p=
0.0
35)
Hyp
os c
onfir
med
by
BG: 6
.0 (4
.9)/
30 d
ays
(vs
base
line
p=
0.03
)
Mon
th: 3
Mea
n: 7
.66%
SD: 0
.71
CSI
I +so
lubl
eM
onth
: 3M
ean:
40.
8 U
/day
SD: 8
.8Si
gnifi
canc
e: p
= n
s
Mon
th: 3
Mea
n: 7
2.8
kgSD
: 9.9
Sign
ifica
nce:
p
= n
s
No.
pre
ferr
ing:
not
rep
orte
d
QoL
: not
rep
orte
d
Skin
rea
ctio
n: 1
5Se
vere
hyp
o: 0
Disc
olou
ratio
n of
insu
lin in
res
ervo
ir or
cat
hete
r (d
idno
t le
ad t
o ob
stru
ctio
n or
sig
nific
ant
hype
rgly
caem
ia):
2
Hyp
os: 1
0.8
(9.9
)/30
day
s (v
s ba
selin
e an
d lis
pro
p=
ns).
Hyp
os c
onfir
med
by
BG: 7
.6 (7
.1)/
30 d
ays
(vs
base
line
p=
ns, v
s lis
pro
p=
ns)
Mon
th: 3
Mea
n: 8
%SD
: 0.8
8Si
gnifi
canc
e:p
=0.
0041
Sign
ifica
nce
is fo
r C
SII v
s M
DI e
xcep
t w
here
sta
ted.
Health Technology Assessment 2004; Vol. 8: No. 43
171
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Appendix 14
Calculations for NHS staff costs
Costs/staff Consultant physician Diabetes specialist nurse Dietitian (assuming (assuming discretionary (assuming G grade nurse; senior dietitian, grade 1; point 3 on salary scale) top of salary scale, top of salary scale,
point 5) point 6)
Annual salary (£) 76,700 26,056 25,145Employer’s National Insurance (£) 8,119 1,994 1,910Employer’s pension contribution (£) 5,262 1,730 1,675Overheadsa 24,320 2,216 2,216Capital overheadsa 4,161 2,263 3,606Total annual costs (£) 118,562 34,259 34,552Working time 41 weeks × 40 h 42 wks × 37.5 h 42 weeks × 37 hCost per hour (£) 72.29 21.75 22.23Cost per day (£) 578.35 163.14 164.53
a Overhead estimates based on data from PSSRU.138
Source: Salary scales from Southampton General Hospital Trust (2001–2002).
Health Technology Assessment 2004; Vol. 8: No. 43
183
Health Technology AssessmentProgramme
Prioritisation Strategy GroupMembers
Chair,Professor Tom Walley, Director, NHS HTA Programme,Department of Pharmacology &Therapeutics,University of Liverpool
Professor Bruce Campbell,Consultant Vascular & GeneralSurgeon, Royal Devon & ExeterHospital
Professor Shah Ebrahim,Professor in Epidemiology of Ageing, University of Bristol
Dr John Reynolds, ClinicalDirector, Acute GeneralMedicine SDU, RadcliffeHospital, Oxford
Dr Ron Zimmern, Director,Public Health Genetics Unit,Strangeways ResearchLaboratories, Cambridge
HTA Commissioning BoardMembers
Programme Director, Professor Tom Walley, Director, NHS HTA Programme,Department of Pharmacology &Therapeutics,University of Liverpool
Chair,Professor Shah Ebrahim,Professor in Epidemiology ofAgeing, Department of SocialMedicine, University of Bristol
Deputy Chair, Professor Jenny Hewison,Professor of Health CarePsychology, Academic Unit ofPsychiatry and BehaviouralSciences, University of LeedsSchool of Medicine
Dr Jeffrey AronsonReader in ClinicalPharmacology, Department ofClinical Pharmacology,Radcliffe Infirmary, Oxford
Professor Ann Bowling,Professor of Health ServicesResearch, Primary Care andPopulation Studies,University College London
Professor Andrew Bradbury,Professor of Vascular Surgery,Department of Vascular Surgery,Birmingham HeartlandsHospital
Professor John Brazier, Directorof Health Economics, Sheffield Health EconomicsGroup, School of Health &Related Research, University of Sheffield
Dr Andrew Briggs, PublicHealth Career Scientist, HealthEconomics Research Centre,University of Oxford
Professor Nicky Cullum,Director of Centre for EvidenceBased Nursing, Department ofHealth Sciences, University ofYork
Dr Andrew Farmer, SeniorLecturer in General Practice,Department of Primary HealthCare, University of Oxford
Professor Fiona J Gilbert,Professor of Radiology,Department of Radiology,University of Aberdeen
Professor Adrian Grant,Director, Health ServicesResearch Unit, University ofAberdeen
Professor F D Richard Hobbs,Professor of Primary Care &General Practice, Department ofPrimary Care & GeneralPractice, University ofBirmingham
Professor Peter Jones, Head ofDepartment, UniversityDepartment of Psychiatry,University of Cambridge
Professor Sallie Lamb, ResearchProfessor in Physiotherapy/Co-Director, InterdisciplinaryResearch Centre in Health,Coventry University
Professor Julian Little,Professor of Epidemiology,Department of Medicine andTherapeutics, University ofAberdeen
Professor Stuart Logan,Director of Health & SocialCare Research, The PeninsulaMedical School, Universities ofExeter & Plymouth
Professor Tim Peters, Professorof Primary Care Health ServicesResearch, Division of PrimaryHealth Care, University ofBristol
Professor Ian Roberts, Professorof Epidemiology & PublicHealth, Intervention ResearchUnit, London School ofHygiene and Tropical Medicine
Professor Peter Sandercock,Professor of Medical Neurology,Department of ClinicalNeurosciences, University ofEdinburgh
Professor Mark Sculpher,Professor of Health Economics,Centre for Health Economics,Institute for Research in theSocial Services, University of York
Professor Martin Severs,Professor in Elderly HealthCare, Portsmouth Institute ofMedicine
Dr Jonathan Shapiro, SeniorFellow, Health ServicesManagement Centre,Birmingham
Ms Kate Thomas,Deputy Director,Medical Care Research Unit,University of Sheffield
Professor Simon G Thompson,Director, MRC BiostatisticsUnit, Institute of Public Health,Cambridge
Ms Sue Ziebland,Senior Research Fellow,Cancer Research UK,University of Oxford
Current and past membership details of all HTA ‘committees’ are available from the HTA website (www.ncchta.org)
© Queen’s Printer and Controller of HMSO 2004. All rights reserved.
Health Technology Assessment Programme
184
Diagnostic Technologies & Screening PanelMembers
Chair,Dr Ron Zimmern, Director ofthe Public Health Genetics Unit,Strangeways ResearchLaboratories, Cambridge
Ms Norma Armston,Freelance Consumer Advocate,Bolton
Professor Max BachmannProfessor Health Care Interfaces, Department of Health Policy and Practice,University of East Anglia
Professor Rudy BilousProfessor of Clinical Medicine &Consultant Physician,The Academic Centre,South Tees Hospitals NHS Trust
Dr Paul Cockcroft, Consultant MedicalMicrobiologist/LaboratoryDirector, Public HealthLaboratory, St Mary’s Hospital, Portsmouth
Professor Adrian K Dixon,Professor of Radiology,Addenbrooke’s Hospital,Cambridge
Dr David Elliman, Consultant in Community Child Health, London
Professor Glyn Elwyn,Primary Medical Care Research Group,Swansea Clinical School,University of WalesSwansea
Dr John Fielding,Consultant Radiologist,Radiology Department,Royal Shrewsbury Hospital
Dr Karen N Foster, ClinicalLecturer, Dept of GeneralPractice & Primary Care,University of Aberdeen
Professor Antony J Franks,Deputy Medical Director, The Leeds Teaching HospitalsNHS Trust
Mr Tam Fry, HonoraryChairman, Child GrowthFoundation, London
Dr Edmund Jessop,Medical Adviser,National SpecialistCommissioning Advisory Group(NSCAG), Department ofHealth, London
Dr Jennifer J Kurinczuk,Consultant ClinicalEpidemiologist,National PerinatalEpidemiology Unit,Oxford
Dr Susanne M Ludgate, MedicalDirector, Medical DevicesAgency, London
Dr William Rosenberg, SeniorLecturer and Consultant inMedicine, University ofSouthampton
Dr Susan Schonfield, CPHMSpecialised ServicesCommissioning, CroydonPrimary Care Trust
Dr Margaret Somerville,Director of Public Health,Teignbridge Primary Care Trust
Professor Lindsay WilsonTurnbull, Scientific Director,Centre for MR Investigations &YCR Professor of Radiology,University of Hull
Professor Martin J Whittle,Head of Division ofReproductive & Child Health,University of Birmingham
Dr Dennis Wright, ConsultantBiochemist & Clinical Director,Pathology & The KennedyGalton Centre, Northwick Park& St Mark’s Hospitals, Harrow
Pharmaceuticals PanelMembers
Chair,Dr John Reynolds, ClinicalDirector, Acute GeneralMedicine SDU, OxfordRadcliffe Hospital
Professor Tony Avery, Professor of Primary HealthCare, University of Nottingham
Professor Stirling Bryan,Professor of Health Economics,Health Services Management Centre,University of Birmingham
Mr Peter Cardy, ChiefExecutive, Macmillan CancerRelief, London
Dr Christopher Cates, GP andCochrane Editor, Bushey HealthCentre
Professor Imti Choonara,Professor in Child Health,University of Nottingham,Derbyshire Children’s Hospital
Mr Charles Dobson, SpecialProjects Adviser, Department ofHealth
Dr Robin Ferner, ConsultantPhysician and Director, WestMidlands Centre for AdverseDrug Reactions, City HospitalNHS Trust, Birmingham
Dr Karen A Fitzgerald,Pharmaceutical Adviser, Bro TafHealth Authority, Cardiff
Mrs Sharon Hart, ManagingEditor, Drug & TherapeuticsBulletin, London
Dr Christine Hine, Consultant inPublic Health Medicine, Bristol South & West PrimaryCare Trust
Professor Stan Kaye,Professor of Medical Oncology,Consultant in MedicalOncology/Drug Development,The Royal Marsden Hospital
Ms Barbara Meredith,Project Manager ClinicalGuidelines, Patient InvolvementUnit, NICE
Dr Frances Rotblat, CPMPDelegate, Medicines ControlAgency, London
Professor Jan Scott,Professor of PsychologicalTreatments,Institute of Psychiatry,University of London
Mrs Katrina Simister, NewProducts Manager, NationalPrescribing Centre, Liverpool
Dr Richard Tiner, MedicalDirector, Association of theBritish Pharmaceutical Industry
Dr Helen Williams,Consultant Microbiologist,Norfolk & Norwich UniversityHospital NHS Trust
Current and past membership details of all HTA ‘committees’ are available from the HTA website (www.ncchta.org)
Health Technology Assessment 2004; Vol. 8: No. 43
185
Therapeutic Procedures PanelMembers
Chair, Professor Bruce Campbell,Consultant Vascular andGeneral Surgeon, Royal Devon& Exeter Hospital
Dr Mahmood Adil, Head ofClinical Support & HealthProtection, Directorate ofHealth and Social Care (North),Department of Health,Manchester
Dr Aileen Clarke,Reader in Health ServicesResearch, Public Health &Policy Research Unit,Barts & the London School ofMedicine & Dentistry,Institute of Community HealthSciences, Queen Mary,University of London
Mr Matthew William Cooke,Senior Clinical Lecturer andHonorary Consultant,Emergency Department,University of Warwick, Coventry& Warwickshire NHS Trust,Division of Health in theCommunity, Centre for PrimaryHealth Care Studies, Coventry
Dr Carl E Counsell, SeniorLecturer in Neurology,University of Aberdeen
Dr Keith Dodd, ConsultantPaediatrician, DerbyshireChildren’s Hospital
Professor Gene Feder, Professorof Primary Care R&D, Barts &the London, Queen Mary’sSchool of Medicine andDentistry, University of London
Professor Paul Gregg,Professor of OrthopaedicSurgical Science, Department ofOrthopaedic Surgery,South Tees Hospital NHS Trust
Ms Bec Hanley, FreelanceConsumer Advocate,Hurstpierpoint
Ms Maryann L. Hardy,Lecturer, Division of Radiography,University of Bradford
Professor Alan Horwich,Director of Clinical R&D, TheInstitute of Cancer Research,London
Dr Phillip Leech, PrincipalMedical Officer for PrimaryCare, Department of Health,London
Dr Simon de Lusignan,Senior Lecturer, Primary CareInformatics, Department ofCommunity Health Sciences,St George’s Hospital MedicalSchool, London
Dr Mike McGovern, SeniorMedical Officer, Heart Team,Department of Health, London
Professor James Neilson,Professor of Obstetrics andGynaecology, Dept of Obstetricsand Gynaecology,University of Liverpool,Liverpool Women’s Hospital
Dr John C Pounsford,Consultant Physician, NorthBristol NHS Trust
Dr Vimal Sharma,Consultant Psychiatrist & HonSnr Lecturer,Mental Health Resource Centre,Victoria Central Hospital,Wirrall
Dr L David Smith, ConsultantCardiologist, Royal Devon &Exeter Hospital
Professor Norman Waugh,Professor of Public Health,University of Aberdeen
Current and past membership details of all HTA ‘committees’ are available from the HTA website (www.ncchta.org)
Health Technology Assessment Programme
186Current and past membership details of all HTA ‘committees’ are available from the HTA website (www.ncchta.org)
Expert Advisory NetworkMembers
Professor Douglas Altman,Director of CSM & CancerResearch UK Med Stat Gp,Centre for Statistics inMedicine, University of Oxford,Institute of Health Sciences,Headington, Oxford
Professor John Bond,Director, Centre for HealthServices Research,University of Newcastle uponTyne, School of Population &Health Sciences,Newcastle upon Tyne
Mr Shaun Brogan, Chief Executive, RidgewayPrimary Care Group, Aylesbury
Mrs Stella Burnside OBE,Chief Executive,Office of the Chief Executive.Trust Headquarters,Altnagelvin Hospitals Health &Social Services Trust,Altnagelvin Area Hospital,Londonderry
Ms Tracy Bury, Project Manager, WorldConfederation for PhysicalTherapy, London
Mr John A Cairns, Professor of Health Economics,Health Economics ResearchUnit, University of Aberdeen
Professor Iain T Cameron,Professor of Obstetrics andGynaecology and Head of theSchool of Medicine,University of Southampton
Dr Christine Clark,Medical Writer & ConsultantPharmacist, Rossendale
Professor Collette Mary Clifford,Professor of Nursing & Head ofResearch, School of HealthSciences, University ofBirmingham, Edgbaston,Birmingham
Professor Barry Cookson,Director, Laboratory of HealthcareAssociated Infection,Health Protection Agency,London
Professor Howard Stephen Cuckle, Professor of ReproductiveEpidemiology, Department ofPaediatrics, Obstetrics &Gynaecology, University ofLeeds
Professor Nicky Cullum, Director of Centre for EvidenceBased Nursing, University of York
Dr Katherine Darton, Information Unit, MIND – TheMental Health Charity, London
Professor Carol Dezateux, Professor of PaediatricEpidemiology, London
Mr John Dunning,Consultant CardiothoracicSurgeon, CardiothoracicSurgical Unit, PapworthHospital NHS Trust, Cambridge
Mr Jonothan Earnshaw,Consultant Vascular Surgeon,Gloucestershire Royal Hospital,Gloucester
Professor Martin Eccles, Professor of ClinicalEffectiveness, Centre for HealthServices Research, University ofNewcastle upon Tyne
Professor Pam Enderby,Professor of CommunityRehabilitation, Institute ofGeneral Practice and PrimaryCare, University of Sheffield
Mr Leonard R Fenwick, Chief Executive, Newcastleupon Tyne Hospitals NHS Trust
Professor David Field, Professor of Neonatal Medicine,Child Health, The LeicesterRoyal Infirmary NHS Trust
Mrs Gillian Fletcher, Antenatal Teacher & Tutor andPresident, National ChildbirthTrust, Henfield
Professor Jayne Franklyn,Professor of Medicine,Department of Medicine,University of Birmingham,Queen Elizabeth Hospital,Edgbaston, Birmingham
Ms Grace Gibbs, Deputy Chief Executive,Director for Nursing, Midwifery& Clinical Support Servs, West Middlesex UniversityHospital, Isleworth
Dr Neville Goodman, Consultant Anaesthetist,Southmead Hospital, Bristol
Professor Alastair Gray,Professor of Health Economics,Department of Public Health,University of Oxford
Professor Robert E Hawkins, CRC Professor and Director ofMedical Oncology, Christie CRCResearch Centre, ChristieHospital NHS Trust, Manchester
Professor F D Richard Hobbs, Professor of Primary Care &General Practice, Department ofPrimary Care & GeneralPractice, University ofBirmingham
Professor Allen Hutchinson, Director of Public Health &Deputy Dean of ScHARR,Department of Public Health,University of Sheffield
Dr Duncan Keeley,General Practitioner (Dr Burch& Ptnrs), The Health Centre,Thame
Dr Donna Lamping,Research Degrees ProgrammeDirector & Reader in Psychology,Health Services Research Unit,London School of Hygiene andTropical Medicine, London
Mr George Levvy,Chief Executive, MotorNeurone Disease Association,Northampton
Professor James Lindesay,Professor of Psychiatry for theElderly, University of Leicester,Leicester General Hospital
Professor Rajan Madhok, Medical Director & Director ofPublic Health, Directorate ofClinical Strategy & PublicHealth, North & East Yorkshire& Northern Lincolnshire HealthAuthority, York
Professor David Mant, Professor of General Practice,Department of Primary Care,University of Oxford
Professor Alexander Markham, Director, Molecular MedicineUnit, St James’s UniversityHospital, Leeds
Dr Chris McCall, General Practitioner, The Hadleigh Practice, Castle Mullen
Professor Alistair McGuire, Professor of Health Economics,London School of Economics
Dr Peter Moore, Freelance Science Writer,Ashtead
Dr Andrew Mortimore, Consultant in Public HealthMedicine, Southampton CityPrimary Care Trust
Dr Sue Moss, Associate Director, CancerScreening Evaluation Unit,Institute of Cancer Research,Sutton
Professor Jon Nicholl, Director of Medical CareResearch Unit, School of Healthand Related Research,University of Sheffield
Mrs Julietta Patnick, National Co-ordinator, NHSCancer Screening Programmes,Sheffield
Professor Robert Peveler,Professor of Liaison Psychiatry,University Mental HealthGroup, Royal South HantsHospital, Southampton
Professor Chris Price, Visiting Chair – Oxford, Clinical Research, BayerDiagnostics Europe, Cirencester
Ms Marianne Rigge, Director, College of Health,London
Dr Eamonn Sheridan,Consultant in Clinical Genetics,Genetics Department,St James’s University Hospital,Leeds
Dr Ken Stein,Senior Clinical Lecturer inPublic Health, Director,Peninsula TechnologyAssessment Group, University of Exeter
Professor Sarah Stewart-Brown, Director HSRU/HonoraryConsultant in PH Medicine,Department of Public Health,University of Oxford
Professor Ala Szczepura, Professor of Health ServiceResearch, Centre for HealthServices Studies, University ofWarwick
Dr Ross Taylor, Senior Lecturer, Department of General Practiceand Primary Care, University of Aberdeen
Mrs Joan Webster, Consumer member, HTA –Expert Advisory Network
How to obtain copies of this and other HTA Programme reports.An electronic version of this publication, in Adobe Acrobat format, is available for downloading free ofcharge for personal use from the HTA website (http://www.ncchta.org). A fully searchable CD-ROM isalso available (see below).
Printed copies of HTA monographs cost £20 each (post and packing free in the UK) to both public andprivate sector purchasers from our Despatch Agents, York Publishing Services.
Non-UK purchasers will have to pay a small fee for post and packing. For European countries the cost is£2 per monograph and for the rest of the world £3 per monograph.
You can order HTA monographs from our Despatch Agents, York Publishing Services by:
– fax (with credit card or official purchase order) – post (with credit card or official purchase order or cheque)– phone during office hours (credit card only).
Additionally the HTA website allows you either to pay securely by credit card or to print out yourorder and then post or fax it.
Contact details are as follows:York Publishing Services Email: ncchta@yps-publishing.co.ukPO Box 642 Tel: 0870 1616662YORK YO31 7WX Fax: 0870 1616663UK Fax from outside the UK: +44 1904 430868
NHS libraries can subscribe free of charge. Public libraries can subscribe at a very reduced cost of £100 for each volume (normally comprising 30–40 titles). The commercial subscription rate is £300 per volume. Please contact York Publishing Services at the address above. Subscriptions can only bepurchased for the current or forthcoming volume.
Payment methods
Paying by chequeIf you pay by cheque, the cheque must be in pounds sterling, made payable to York PublishingDistribution and drawn on a bank with a UK address.
Paying by credit cardThe following cards are accepted by phone, fax, post or via the website ordering pages: Delta, Eurocard,Mastercard, Solo, Switch and Visa. We advise against sending credit card details in a plain email.
Paying by official purchase orderYou can post or fax these, but they must be from public bodies (i.e. NHS or universities) within the UK.We cannot at present accept purchase orders from commercial companies or from outside the UK.
How do I get a copy of HTA on CD?
Please use the form on the HTA website (www.ncchta.org/htacd.htm). Or contact York PublishingServices (see contact details above) by email, post, fax or phone. HTA on CD is currently free of chargeworldwide.
The website also provides information about the HTA Programme and lists the membership of the variouscommittees.
HTA
Health Technology Assessm
ent 2004;Vol. 8: No. 43
Clinical and cost-effectiveness of continuous subcutaneous insulin infusion for diabetes
Clinical and cost-effectiveness of continuous subcutaneous insulin infusion for diabetes
JL Colquitt, C Green, MK Sidhu, D Hartwell and N Waugh
Health Technology Assessment 2004; Vol. 8: No. 43
HTAHealth Technology AssessmentNHS R&D HTA Programme
The National Coordinating Centre for Health Technology Assessment,Mailpoint 728, Boldrewood,University of Southampton,Southampton, SO16 7PX, UK.Fax: +44 (0) 23 8059 5639 Email: hta@soton.ac.ukhttp://www.ncchta.org ISSN 1366-5278
FeedbackThe HTA Programme and the authors would like to know
your views about this report.
The Correspondence Page on the HTA website(http://www.ncchta.org) is a convenient way to publish
your comments. If you prefer, you can send your comments to the address below, telling us whether you would like
us to transfer them to the website.
We look forward to hearing from you.
October 2004
top related